Hugging Face's logo

Datasets:
Intelligent-Internet
/
pd12m

Tasks:
Feature Extraction
Modalities:
Image
Tabular
Text
Formats:
csv
Languages:
English
Size:
10M - 100M
ArXiv:
Libraries:
Datasets
Dask
License:
Dataset card Data Studio Files
xet
Community
4
Leask commited on
Commit
5d7a4a1
·
verified ·
1 Parent(s): f1da9bf

Add files using upload-large-folder tool

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .history/README_20250513101818.md +109 -0
  2. .history/README_20250513101823.md +109 -0
  3. .history/README_20250513101834.md +109 -0
  4. .history/README_20250513101836.md +109 -0
  5. .history/README_20250513101846.md +109 -0
  6. .history/README_20250513101854.md +109 -0
  7. .history/README_20250513102007.md +109 -0
  8. .history/README_20250513103952.md +109 -0
  9. .history/README_20250513103953.md +109 -0
  10. .history/README_20250513103954.md +109 -0
  11. .history/README_20250513103957.md +109 -0
  12. .history/README_20250513104030.md +109 -0
  13. .history/README_20250513104104.md +109 -0
  14. .history/README_20250513104204.md +109 -0
  15. .history/README_20250513104214.md +109 -0
  16. .history/README_20250513104240.md +109 -0
  17. .history/README_20250513111019.md +108 -0
  18. .history/README_20250513111025.md +109 -0
  19. .history/README_20250513111030.md +109 -0
  20. .history/README_20250513111043.md +109 -0
  21. .history/README_20250513111111.md +109 -0
  22. .history/README_20250513111112.md +111 -0
  23. .history/README_20250513112320.md +162 -0
  24. .history/README_20250513112324.md +162 -0
  25. .history/README_20250513112329.md +162 -0
  26. .history/README_20250513112334.md +162 -0
  27. .history/README_20250513112341.md +164 -0
  28. .history/README_20250513112353.md +164 -0
  29. .history/README_20250513112357.md +164 -0
  30. .history/README_20250513112401.md +164 -0
  31. .history/README_20250513112411.md +164 -0
  32. .history/README_20250513112420.md +164 -0
  33. .history/README_20250513112423.md +164 -0
  34. .history/README_20250513112429.md +164 -0
  35. .history/README_20250513112431.md +164 -0
  36. .history/README_20250513112436.md +164 -0
  37. .history/README_20250513112439.md +164 -0
  38. .history/README_20250513112452.md +172 -0
  39. .history/README_20250513112515.md +169 -0
  40. .history/README_20250513112522.md +170 -0
  41. .history/README_20250513112531.md +168 -0
  42. .history/README_20250513112617.md +168 -0
  43. .history/README_20250513112619.md +170 -0
  44. .history/README_20250513112620.md +168 -0
  45. .history/README_20250513112719.md +165 -0
  46. .history/README_20250513112725.md +171 -0
  47. .history/README_20250513112727.md +170 -0
  48. .history/README_20250513112728.md +176 -0
  49. .history/README_20250513112733.md +176 -0
  50. .history/README_20250513122557.md +128 -0
.history/README_20250513101818.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated PD12M image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using SigLIP2. All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
53
+ meta JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS {table_name}_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ f"CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ f"CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ f'CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513101823.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated PD12M image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using SigLIP2. All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
53
+ meta JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS {table_name}_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ f"CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ f"CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ f'CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513101834.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated PD12M image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using SigLIP2. All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
53
+ meta JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513101836.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated PD12M image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using SigLIP2. All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
53
+ meta JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513101846.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated PD12M image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using SigLIP2. All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
53
+ meta JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513101854.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated PD12M image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using SigLIP2. All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
53
+ meta JSONB NOT NULL DEFAULT '{EMPTY_OBJECT}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513102007.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated PD12M image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using SigLIP2. All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513103952.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated PD12M image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using SigLIP 2】. All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513103953.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated PD12M image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using SigLIP 2]. All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513103954.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated PD12M image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2]. All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513103957.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated PD12M image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513104030.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [PD12M] image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513104104.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [PD12M](https://huggingface.co/datasets/Spawning/PD12M) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513104204.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://huggingface.co/datasets/Spawning/PD12M) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513104214.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [PD12M](https://huggingface.co/datasets/Spawning/PD12M) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513104240.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513111019.md ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+ 1. Load the dataset from local file system to a remote PostgreSQL server:
63
+
64
+ ```sql
65
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
66
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
68
+ ...
69
+ ```
70
+
71
+ 2.2. Load the dataset from the PostgreSQL server's file system:
72
+
73
+ ```sql
74
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
75
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
77
+ ...
78
+ ```
79
+
80
+ 3. Create Indexes
81
+
82
+ ```sql
83
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
84
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
97
+ residual_quantization = true
98
+ [build.internal]
99
+ lists = [20000]
100
+ build_threads = 6
101
+ spherical_centroids = false
102
+ $$);
103
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
106
+ ```
107
+
108
+ ## Dataset Structure
.history/README_20250513111025.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513111030.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513111043.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Uses
26
+
27
+ This dataset is available for a wide range of applications.
28
+
29
+ ### Use with Chipmunk
30
+
31
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
32
+
33
+ ### Use directly in PostgreSQL
34
+
35
+ 1. Create Table
36
+
37
+ ```sql
38
+ CREATE TABLE IF NOT EXISTS is_pd12m (
39
+ id BIGSERIAL PRIMARY KEY,
40
+ url VARCHAR NOT NULL,
41
+ hash VARCHAR NOT NULL DEFAULT '',
42
+ caption VARCHAR NOT NULL DEFAULT '',
43
+ caption_long VARCHAR NOT NULL DEFAULT '',
44
+ origin_hash VARCHAR NOT NULL DEFAULT '',
45
+ origin_width BIGINT NOT NULL DEFAULT 0,
46
+ origin_height BIGINT NOT NULL DEFAULT 0,
47
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
48
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
49
+ processed_width BIGINT NOT NULL DEFAULT 0,
50
+ processed_height BIGINT NOT NULL DEFAULT 0,
51
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
52
+ exif JSONB NOT NULL DEFAULT '{}',
53
+ meta JSONB NOT NULL DEFAULT '{}',
54
+ source JSONB NOT NULL DEFAULT '[]',
55
+ vector halfvec(1152) DEFAULT NULL,
56
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
57
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
58
+ );
59
+ ```
60
+
61
+ 2. Load csv file to database
62
+
63
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
64
+
65
+ ```sql
66
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
67
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
68
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
69
+ ...
70
+ ```
71
+
72
+ 2.2. Load the dataset from the PostgreSQL server's file system:
73
+
74
+ ```sql
75
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
76
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
77
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
78
+ ...
79
+ ```
80
+
81
+ 3. Create Indexes
82
+
83
+ ```sql
84
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
85
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
86
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
98
+ residual_quantization = true
99
+ [build.internal]
100
+ lists = [20000]
101
+ build_threads = 6
102
+ spherical_centroids = false
103
+ $$);
104
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
105
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
107
+ ```
108
+
109
+ ## Dataset Structure
.history/README_20250513111111.md ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ ## Uses
28
+
29
+ This dataset is available for a wide range of applications.
30
+
31
+ ### Use with Chipmunk
32
+
33
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
34
+
35
+ ### Use directly in PostgreSQL
36
+
37
+ 1. Create Table
38
+
39
+ ```sql
40
+ CREATE TABLE IF NOT EXISTS is_pd12m (
41
+ id BIGSERIAL PRIMARY KEY,
42
+ url VARCHAR NOT NULL,
43
+ hash VARCHAR NOT NULL DEFAULT '',
44
+ caption VARCHAR NOT NULL DEFAULT '',
45
+ caption_long VARCHAR NOT NULL DEFAULT '',
46
+ origin_hash VARCHAR NOT NULL DEFAULT '',
47
+ origin_width BIGINT NOT NULL DEFAULT 0,
48
+ origin_height BIGINT NOT NULL DEFAULT 0,
49
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
50
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
51
+ processed_width BIGINT NOT NULL DEFAULT 0,
52
+ processed_height BIGINT NOT NULL DEFAULT 0,
53
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
54
+ exif JSONB NOT NULL DEFAULT '{}',
55
+ meta JSONB NOT NULL DEFAULT '{}',
56
+ source JSONB NOT NULL DEFAULT '[]',
57
+ vector halfvec(1152) DEFAULT NULL,
58
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
59
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
60
+ );
61
+ ```
62
+
63
+ 2. Load csv file to database
64
+
65
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
66
+
67
+ ```sql
68
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
69
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
70
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
71
+ ...
72
+ ```
73
+
74
+ 2.2. Load the dataset from the PostgreSQL server's file system:
75
+
76
+ ```sql
77
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
78
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
79
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
80
+ ...
81
+ ```
82
+
83
+ 3. Create Indexes
84
+
85
+ ```sql
86
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
87
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
88
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
98
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
99
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
100
+ residual_quantization = true
101
+ [build.internal]
102
+ lists = [20000]
103
+ build_threads = 6
104
+ spherical_centroids = false
105
+ $$);
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
107
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
108
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
109
+ ```
.history/README_20250513111112.md ADDED
@@ -0,0 +1,111 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ## Uses
30
+
31
+ This dataset is available for a wide range of applications.
32
+
33
+ ### Use with Chipmunk
34
+
35
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
36
+
37
+ ### Use directly in PostgreSQL
38
+
39
+ 1. Create Table
40
+
41
+ ```sql
42
+ CREATE TABLE IF NOT EXISTS is_pd12m (
43
+ id BIGSERIAL PRIMARY KEY,
44
+ url VARCHAR NOT NULL,
45
+ hash VARCHAR NOT NULL DEFAULT '',
46
+ caption VARCHAR NOT NULL DEFAULT '',
47
+ caption_long VARCHAR NOT NULL DEFAULT '',
48
+ origin_hash VARCHAR NOT NULL DEFAULT '',
49
+ origin_width BIGINT NOT NULL DEFAULT 0,
50
+ origin_height BIGINT NOT NULL DEFAULT 0,
51
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
52
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
53
+ processed_width BIGINT NOT NULL DEFAULT 0,
54
+ processed_height BIGINT NOT NULL DEFAULT 0,
55
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
56
+ exif JSONB NOT NULL DEFAULT '{}',
57
+ meta JSONB NOT NULL DEFAULT '{}',
58
+ source JSONB NOT NULL DEFAULT '[]',
59
+ vector halfvec(1152) DEFAULT NULL,
60
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
61
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
62
+ );
63
+ ```
64
+
65
+ 2. Load csv file to database
66
+
67
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
68
+
69
+ ```sql
70
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
71
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
72
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
73
+ ...
74
+ ```
75
+
76
+ 2.2. Load the dataset from the PostgreSQL server's file system:
77
+
78
+ ```sql
79
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
80
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
81
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
82
+ ...
83
+ ```
84
+
85
+ 3. Create Indexes
86
+
87
+ ```sql
88
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
89
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
90
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
91
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
92
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
93
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
94
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
95
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
96
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
97
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
98
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
99
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
100
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
101
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
102
+ residual_quantization = true
103
+ [build.internal]
104
+ lists = [20000]
105
+ build_threads = 6
106
+ spherical_centroids = false
107
+ $$);
108
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
109
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
110
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
111
+ ```
.history/README_20250513112320.md ADDED
@@ -0,0 +1,162 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+
137
+
138
+ ```sql
139
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
140
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
141
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
153
+ residual_quantization = true
154
+ [build.internal]
155
+ lists = [20000]
156
+ build_threads = 6
157
+ spherical_centroids = false
158
+ $$);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
160
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
162
+ ```
.history/README_20250513112324.md ADDED
@@ -0,0 +1,162 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+
137
+
138
+ ```sql
139
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
140
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
141
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
153
+ residual_quantization = true
154
+ [build.internal]
155
+ lists = [20000]
156
+ build_threads = 6
157
+ spherical_centroids = false
158
+ $$);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
160
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
162
+ ```
.history/README_20250513112329.md ADDED
@@ -0,0 +1,162 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+
137
+
138
+ ```sql
139
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
140
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
141
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
153
+ residual_quantization = true
154
+ [build.internal]
155
+ lists = [20000]
156
+ build_threads = 6
157
+ spherical_centroids = false
158
+ $$);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
160
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
162
+ ```
.history/README_20250513112334.md ADDED
@@ -0,0 +1,162 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes:
137
+
138
+ ```sql
139
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
140
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
141
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
153
+ residual_quantization = true
154
+ [build.internal]
155
+ lists = [20000]
156
+ build_threads = 6
157
+ spherical_centroids = false
158
+ $$);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
160
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
162
+ ```
.history/README_20250513112341.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes for the best performance:
137
+
138
+
139
+
140
+ ```sql
141
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
155
+ residual_quantization = true
156
+ [build.internal]
157
+ lists = [20000]
158
+ build_threads = 6
159
+ spherical_centroids = false
160
+ $$);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
164
+ ```
.history/README_20250513112353.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes for the best performance.
137
+
138
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
139
+
140
+ ```sql
141
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
155
+ residual_quantization = true
156
+ [build.internal]
157
+ lists = [20000]
158
+ build_threads = 6
159
+ spherical_centroids = false
160
+ $$);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
164
+ ```
.history/README_20250513112357.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes for the best performance.
137
+
138
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can
139
+
140
+ ```sql
141
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
155
+ residual_quantization = true
156
+ [build.internal]
157
+ lists = [20000]
158
+ build_threads = 6
159
+ spherical_centroids = false
160
+ $$);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
164
+ ```
.history/README_20250513112401.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes for the best performance.
137
+
138
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more
139
+
140
+ ```sql
141
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
155
+ residual_quantization = true
156
+ [build.internal]
157
+ lists = [20000]
158
+ build_threads = 6
159
+ spherical_centroids = false
160
+ $$);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
164
+ ```
.history/README_20250513112411.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes for the best performance.
137
+
138
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about
139
+
140
+ ```sql
141
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
155
+ residual_quantization = true
156
+ [build.internal]
157
+ lists = [20000]
158
+ build_threads = 6
159
+ spherical_centroids = false
160
+ $$);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
164
+ ```
.history/README_20250513112420.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes for the best performance.
137
+
138
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from
139
+
140
+ ```sql
141
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
155
+ residual_quantization = true
156
+ [build.internal]
157
+ lists = [20000]
158
+ build_threads = 6
159
+ spherical_centroids = false
160
+ $$);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
164
+ ```
.history/README_20250513112423.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes for the best performance.
137
+
138
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the vec
139
+
140
+ ```sql
141
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
155
+ residual_quantization = true
156
+ [build.internal]
157
+ lists = [20000]
158
+ build_threads = 6
159
+ spherical_centroids = false
160
+ $$);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
164
+ ```
.history/README_20250513112429.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes for the best performance.
137
+
138
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the vector
139
+
140
+ ```sql
141
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
155
+ residual_quantization = true
156
+ [build.internal]
157
+ lists = [20000]
158
+ build_threads = 6
159
+ spherical_centroids = false
160
+ $$);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
164
+ ```
.history/README_20250513112431.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes for the best performance.
137
+
138
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://github.com/tensorchord/vectorchord) repository.
139
+
140
+ ```sql
141
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
155
+ residual_quantization = true
156
+ [build.internal]
157
+ lists = [20000]
158
+ build_threads = 6
159
+ spherical_centroids = false
160
+ $$);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
164
+ ```
.history/README_20250513112436.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes for the best performance.
137
+
138
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://github.com/tensorchord/vectorchord) documentation.
139
+
140
+ ```sql
141
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
155
+ residual_quantization = true
156
+ [build.internal]
157
+ lists = [20000]
158
+ build_threads = 6
159
+ spherical_centroids = false
160
+ $$);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
164
+ ```
.history/README_20250513112439.md ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+
28
+
29
+ ```
30
+ \d is_pd12m;
31
+ Table "public.is_pd12m"
32
+ Column | Type | Collation | Nullable | Default
33
+ ----------------------+-----------------------------+-----------+----------+---------------
34
+ id | bigint | | not null |
35
+ url | character varying | | not null |
36
+ caption | character varying | | not null |
37
+ caption_long | character varying | | not null |
38
+ origin_width | bigint | | not null |
39
+ origin_height | bigint | | not null |
40
+ origin_storage_id | character varying(1024) | | not null |
41
+ processed_storage_id | character varying(1024) | | not null |
42
+ processed_width | bigint | | not null |
43
+ processed_height | bigint | | not null |
44
+ aspect_ratio | double precision | | not null |
45
+ exif | jsonb | | not null |
46
+ meta | jsonb | | not null |
47
+ created_at | timestamp without time zone | | not null |
48
+ updated_at | timestamp without time zone | | not null |
49
+ source | jsonb | | not null |
50
+ vector | halfvec(1152) | | | NULL::halfvec
51
+ Indexes:
52
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
53
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
54
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
55
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
56
+ "is_pd12m_created_at_index" btree (created_at)
57
+ "is_pd12m_exif_index" gin (exif)
58
+ "is_pd12m_meta_index" gin (meta)
59
+ "is_pd12m_origin_height_index" btree (origin_height)
60
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
61
+ "is_pd12m_origin_width_index" btree (origin_width)
62
+ "is_pd12m_processed_height_index" btree (processed_height)
63
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
64
+ "is_pd12m_processed_width_index" btree (processed_width)
65
+ "is_pd12m_source_index" gin (source)
66
+ "is_pd12m_updated_at_index" btree (updated_at)
67
+ "is_pd12m_url_index" UNIQUE, btree (url)
68
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
69
+ residual_quantization = true
70
+ [build.internal]
71
+ lists = [20000]
72
+ build_threads = 6
73
+ spherical_centroids = false
74
+ ')
75
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
76
+ ```
77
+
78
+ ## Uses
79
+
80
+ This dataset is available for a wide range of applications.
81
+
82
+ ### Use with Chipmunk
83
+
84
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
85
+
86
+ ### Use directly in PostgreSQL
87
+
88
+ 1. Create Table
89
+
90
+ ```sql
91
+ CREATE TABLE IF NOT EXISTS is_pd12m (
92
+ id BIGSERIAL PRIMARY KEY,
93
+ url VARCHAR NOT NULL,
94
+ hash VARCHAR NOT NULL DEFAULT '',
95
+ caption VARCHAR NOT NULL DEFAULT '',
96
+ caption_long VARCHAR NOT NULL DEFAULT '',
97
+ origin_hash VARCHAR NOT NULL DEFAULT '',
98
+ origin_width BIGINT NOT NULL DEFAULT 0,
99
+ origin_height BIGINT NOT NULL DEFAULT 0,
100
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
101
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
102
+ processed_width BIGINT NOT NULL DEFAULT 0,
103
+ processed_height BIGINT NOT NULL DEFAULT 0,
104
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
105
+ exif JSONB NOT NULL DEFAULT '{}',
106
+ meta JSONB NOT NULL DEFAULT '{}',
107
+ source JSONB NOT NULL DEFAULT '[]',
108
+ vector halfvec(1152) DEFAULT NULL,
109
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
110
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
111
+ );
112
+ ```
113
+
114
+ 2. Load csv file to database
115
+
116
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
117
+
118
+ ```sql
119
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
120
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
122
+ ...
123
+ ```
124
+
125
+ 2.2. Load the dataset from the PostgreSQL server's file system:
126
+
127
+ ```sql
128
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
129
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
131
+ ...
132
+ ```
133
+
134
+ 3. Create Indexes
135
+
136
+ You need to create the following indexes for the best performance.
137
+
138
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
139
+
140
+ ```sql
141
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
142
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
155
+ residual_quantization = true
156
+ [build.internal]
157
+ lists = [20000]
158
+ build_threads = 6
159
+ spherical_centroids = false
160
+ $$);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
164
+ ```
.history/README_20250513112452.md ADDED
@@ -0,0 +1,172 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ id: A unique identifier for the image.
28
+ url: The URL of the image.
29
+ caption: A caption for the image.
30
+ width: The width of the image in pixels.
31
+ height: The height of the image in pixels.
32
+ mime_type: The MIME type of the image file.
33
+ hash: The MD5 hash of the image file.
34
+ license: The URL of the image license.
35
+ source : The source organization of the image.
36
+
37
+ ```
38
+ \d is_pd12m;
39
+ Table "public.is_pd12m"
40
+ Column | Type | Collation | Nullable | Default
41
+ ----------------------+-----------------------------+-----------+----------+---------------
42
+ id | bigint | | not null |
43
+ url | character varying | | not null |
44
+ caption | character varying | | not null |
45
+ caption_long | character varying | | not null |
46
+ origin_width | bigint | | not null |
47
+ origin_height | bigint | | not null |
48
+ origin_storage_id | character varying(1024) | | not null |
49
+ processed_storage_id | character varying(1024) | | not null |
50
+ processed_width | bigint | | not null |
51
+ processed_height | bigint | | not null |
52
+ aspect_ratio | double precision | | not null |
53
+ exif | jsonb | | not null |
54
+ meta | jsonb | | not null |
55
+ created_at | timestamp without time zone | | not null |
56
+ updated_at | timestamp without time zone | | not null |
57
+ source | jsonb | | not null |
58
+ vector | halfvec(1152) | | | NULL::halfvec
59
+ Indexes:
60
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
61
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
62
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
63
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
64
+ "is_pd12m_created_at_index" btree (created_at)
65
+ "is_pd12m_exif_index" gin (exif)
66
+ "is_pd12m_meta_index" gin (meta)
67
+ "is_pd12m_origin_height_index" btree (origin_height)
68
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
69
+ "is_pd12m_origin_width_index" btree (origin_width)
70
+ "is_pd12m_processed_height_index" btree (processed_height)
71
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
72
+ "is_pd12m_processed_width_index" btree (processed_width)
73
+ "is_pd12m_source_index" gin (source)
74
+ "is_pd12m_updated_at_index" btree (updated_at)
75
+ "is_pd12m_url_index" UNIQUE, btree (url)
76
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
77
+ residual_quantization = true
78
+ [build.internal]
79
+ lists = [20000]
80
+ build_threads = 6
81
+ spherical_centroids = false
82
+ ')
83
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
84
+ ```
85
+
86
+ ## Uses
87
+
88
+ This dataset is available for a wide range of applications.
89
+
90
+ ### Use with Chipmunk
91
+
92
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
93
+
94
+ ### Use directly in PostgreSQL
95
+
96
+ 1. Create Table
97
+
98
+ ```sql
99
+ CREATE TABLE IF NOT EXISTS is_pd12m (
100
+ id BIGSERIAL PRIMARY KEY,
101
+ url VARCHAR NOT NULL,
102
+ hash VARCHAR NOT NULL DEFAULT '',
103
+ caption VARCHAR NOT NULL DEFAULT '',
104
+ caption_long VARCHAR NOT NULL DEFAULT '',
105
+ origin_hash VARCHAR NOT NULL DEFAULT '',
106
+ origin_width BIGINT NOT NULL DEFAULT 0,
107
+ origin_height BIGINT NOT NULL DEFAULT 0,
108
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
109
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
110
+ processed_width BIGINT NOT NULL DEFAULT 0,
111
+ processed_height BIGINT NOT NULL DEFAULT 0,
112
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
113
+ exif JSONB NOT NULL DEFAULT '{}',
114
+ meta JSONB NOT NULL DEFAULT '{}',
115
+ source JSONB NOT NULL DEFAULT '[]',
116
+ vector halfvec(1152) DEFAULT NULL,
117
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
118
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
119
+ );
120
+ ```
121
+
122
+ 2. Load csv file to database
123
+
124
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
125
+
126
+ ```sql
127
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
128
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
129
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
130
+ ...
131
+ ```
132
+
133
+ 2.2. Load the dataset from the PostgreSQL server's file system:
134
+
135
+ ```sql
136
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
137
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
138
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
139
+ ...
140
+ ```
141
+
142
+ 3. Create Indexes
143
+
144
+ You need to create the following indexes for the best performance.
145
+
146
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
147
+
148
+ ```sql
149
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
156
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
157
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
158
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
160
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
163
+ residual_quantization = true
164
+ [build.internal]
165
+ lists = [20000]
166
+ build_threads = 6
167
+ spherical_centroids = false
168
+ $$);
169
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
170
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
171
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
172
+ ```
.history/README_20250513112515.md ADDED
@@ -0,0 +1,169 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - width: The width of the image in pixels.
31
+ - height: The height of the image in pixels.
32
+ - mime_type: The MIME type of the image file.
33
+ - hash: The MD5 hash of the image file.
34
+ - license: The URL of the image license.
35
+ - source : The source organization of the image.
36
+
37
+ ```
38
+ \d is_pd12m;
39
+ Table "public.is_pd12m"
40
+ Column | Type | Collation | Nullable | Default
41
+ ----------------------+-----------------------------+-----------+----------+---------------
42
+ caption_long | character varying | | not null |
43
+ origin_width | bigint | | not null |
44
+ origin_height | bigint | | not null |
45
+ origin_storage_id | character varying(1024) | | not null |
46
+ processed_storage_id | character varying(1024) | | not null |
47
+ processed_width | bigint | | not null |
48
+ processed_height | bigint | | not null |
49
+ aspect_ratio | double precision | | not null |
50
+ exif | jsonb | | not null |
51
+ meta | jsonb | | not null |
52
+ created_at | timestamp without time zone | | not null |
53
+ updated_at | timestamp without time zone | | not null |
54
+ source | jsonb | | not null |
55
+ vector | halfvec(1152) | | | NULL::halfvec
56
+ Indexes:
57
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
58
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
59
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
60
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
61
+ "is_pd12m_created_at_index" btree (created_at)
62
+ "is_pd12m_exif_index" gin (exif)
63
+ "is_pd12m_meta_index" gin (meta)
64
+ "is_pd12m_origin_height_index" btree (origin_height)
65
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
66
+ "is_pd12m_origin_width_index" btree (origin_width)
67
+ "is_pd12m_processed_height_index" btree (processed_height)
68
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
69
+ "is_pd12m_processed_width_index" btree (processed_width)
70
+ "is_pd12m_source_index" gin (source)
71
+ "is_pd12m_updated_at_index" btree (updated_at)
72
+ "is_pd12m_url_index" UNIQUE, btree (url)
73
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
74
+ residual_quantization = true
75
+ [build.internal]
76
+ lists = [20000]
77
+ build_threads = 6
78
+ spherical_centroids = false
79
+ ')
80
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
81
+ ```
82
+
83
+ ## Uses
84
+
85
+ This dataset is available for a wide range of applications.
86
+
87
+ ### Use with Chipmunk
88
+
89
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
90
+
91
+ ### Use directly in PostgreSQL
92
+
93
+ 1. Create Table
94
+
95
+ ```sql
96
+ CREATE TABLE IF NOT EXISTS is_pd12m (
97
+ id BIGSERIAL PRIMARY KEY,
98
+ url VARCHAR NOT NULL,
99
+ hash VARCHAR NOT NULL DEFAULT '',
100
+ caption VARCHAR NOT NULL DEFAULT '',
101
+ caption_long VARCHAR NOT NULL DEFAULT '',
102
+ origin_hash VARCHAR NOT NULL DEFAULT '',
103
+ origin_width BIGINT NOT NULL DEFAULT 0,
104
+ origin_height BIGINT NOT NULL DEFAULT 0,
105
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
106
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
107
+ processed_width BIGINT NOT NULL DEFAULT 0,
108
+ processed_height BIGINT NOT NULL DEFAULT 0,
109
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
110
+ exif JSONB NOT NULL DEFAULT '{}',
111
+ meta JSONB NOT NULL DEFAULT '{}',
112
+ source JSONB NOT NULL DEFAULT '[]',
113
+ vector halfvec(1152) DEFAULT NULL,
114
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
115
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
116
+ );
117
+ ```
118
+
119
+ 2. Load csv file to database
120
+
121
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
122
+
123
+ ```sql
124
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
125
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
126
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
127
+ ...
128
+ ```
129
+
130
+ 2.2. Load the dataset from the PostgreSQL server's file system:
131
+
132
+ ```sql
133
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
134
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
135
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
136
+ ...
137
+ ```
138
+
139
+ 3. Create Indexes
140
+
141
+ You need to create the following indexes for the best performance.
142
+
143
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
144
+
145
+ ```sql
146
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
156
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
157
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
158
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
160
+ residual_quantization = true
161
+ [build.internal]
162
+ lists = [20000]
163
+ build_threads = 6
164
+ spherical_centroids = false
165
+ $$);
166
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
167
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
168
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
169
+ ```
.history/README_20250513112522.md ADDED
@@ -0,0 +1,170 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - caption_long: A long caption for the image.
31
+ - width: The width of the image in pixels.
32
+ - height: The height of the image in pixels.
33
+ - mime_type: The MIME type of the image file.
34
+ - hash: The MD5 hash of the image file.
35
+ - license: The URL of the image license.
36
+ - source : The source organization of the image.
37
+
38
+ ```
39
+ \d is_pd12m;
40
+ Table "public.is_pd12m"
41
+ Column | Type | Collation | Nullable | Default
42
+ ----------------------+-----------------------------+-----------+----------+---------------
43
+ caption_long | character varying | | not null |
44
+ origin_width | bigint | | not null |
45
+ origin_height | bigint | | not null |
46
+ origin_storage_id | character varying(1024) | | not null |
47
+ processed_storage_id | character varying(1024) | | not null |
48
+ processed_width | bigint | | not null |
49
+ processed_height | bigint | | not null |
50
+ aspect_ratio | double precision | | not null |
51
+ exif | jsonb | | not null |
52
+ meta | jsonb | | not null |
53
+ created_at | timestamp without time zone | | not null |
54
+ updated_at | timestamp without time zone | | not null |
55
+ source | jsonb | | not null |
56
+ vector | halfvec(1152) | | | NULL::halfvec
57
+ Indexes:
58
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
59
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
60
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
61
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
62
+ "is_pd12m_created_at_index" btree (created_at)
63
+ "is_pd12m_exif_index" gin (exif)
64
+ "is_pd12m_meta_index" gin (meta)
65
+ "is_pd12m_origin_height_index" btree (origin_height)
66
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
67
+ "is_pd12m_origin_width_index" btree (origin_width)
68
+ "is_pd12m_processed_height_index" btree (processed_height)
69
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
70
+ "is_pd12m_processed_width_index" btree (processed_width)
71
+ "is_pd12m_source_index" gin (source)
72
+ "is_pd12m_updated_at_index" btree (updated_at)
73
+ "is_pd12m_url_index" UNIQUE, btree (url)
74
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
75
+ residual_quantization = true
76
+ [build.internal]
77
+ lists = [20000]
78
+ build_threads = 6
79
+ spherical_centroids = false
80
+ ')
81
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
82
+ ```
83
+
84
+ ## Uses
85
+
86
+ This dataset is available for a wide range of applications.
87
+
88
+ ### Use with Chipmunk
89
+
90
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
91
+
92
+ ### Use directly in PostgreSQL
93
+
94
+ 1. Create Table
95
+
96
+ ```sql
97
+ CREATE TABLE IF NOT EXISTS is_pd12m (
98
+ id BIGSERIAL PRIMARY KEY,
99
+ url VARCHAR NOT NULL,
100
+ hash VARCHAR NOT NULL DEFAULT '',
101
+ caption VARCHAR NOT NULL DEFAULT '',
102
+ caption_long VARCHAR NOT NULL DEFAULT '',
103
+ origin_hash VARCHAR NOT NULL DEFAULT '',
104
+ origin_width BIGINT NOT NULL DEFAULT 0,
105
+ origin_height BIGINT NOT NULL DEFAULT 0,
106
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
107
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
108
+ processed_width BIGINT NOT NULL DEFAULT 0,
109
+ processed_height BIGINT NOT NULL DEFAULT 0,
110
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
111
+ exif JSONB NOT NULL DEFAULT '{}',
112
+ meta JSONB NOT NULL DEFAULT '{}',
113
+ source JSONB NOT NULL DEFAULT '[]',
114
+ vector halfvec(1152) DEFAULT NULL,
115
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
116
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
117
+ );
118
+ ```
119
+
120
+ 2. Load csv file to database
121
+
122
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
123
+
124
+ ```sql
125
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
126
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
127
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
128
+ ...
129
+ ```
130
+
131
+ 2.2. Load the dataset from the PostgreSQL server's file system:
132
+
133
+ ```sql
134
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
135
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
136
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
137
+ ...
138
+ ```
139
+
140
+ 3. Create Indexes
141
+
142
+ You need to create the following indexes for the best performance.
143
+
144
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
145
+
146
+ ```sql
147
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
156
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
157
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
158
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
160
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
161
+ residual_quantization = true
162
+ [build.internal]
163
+ lists = [20000]
164
+ build_threads = 6
165
+ spherical_centroids = false
166
+ $$);
167
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
168
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
169
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
170
+ ```
.history/README_20250513112531.md ADDED
@@ -0,0 +1,168 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - caption_long: A long caption for the image.
31
+ - width: The width of the image in pixels.
32
+ - height: The height of the image in pixels.
33
+ - mime_type: The MIME type of the image file.
34
+ - hash: The MD5 hash of the image file.
35
+ - license: The URL of the image license.
36
+ - source : The source organization of the image.
37
+
38
+ ```
39
+ \d is_pd12m;
40
+ Table "public.is_pd12m"
41
+ Column | Type | Collation | Nullable | Default
42
+ origin_width | bigint | | not null |
43
+ origin_height | bigint | | not null |
44
+ origin_storage_id | character varying(1024) | | not null |
45
+ processed_storage_id | character varying(1024) | | not null |
46
+ processed_width | bigint | | not null |
47
+ processed_height | bigint | | not null |
48
+ aspect_ratio | double precision | | not null |
49
+ exif | jsonb | | not null |
50
+ meta | jsonb | | not null |
51
+ created_at | timestamp without time zone | | not null |
52
+ updated_at | timestamp without time zone | | not null |
53
+ source | jsonb | | not null |
54
+ vector | halfvec(1152) | | | NULL::halfvec
55
+ Indexes:
56
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
57
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
58
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
59
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
60
+ "is_pd12m_created_at_index" btree (created_at)
61
+ "is_pd12m_exif_index" gin (exif)
62
+ "is_pd12m_meta_index" gin (meta)
63
+ "is_pd12m_origin_height_index" btree (origin_height)
64
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
65
+ "is_pd12m_origin_width_index" btree (origin_width)
66
+ "is_pd12m_processed_height_index" btree (processed_height)
67
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
68
+ "is_pd12m_processed_width_index" btree (processed_width)
69
+ "is_pd12m_source_index" gin (source)
70
+ "is_pd12m_updated_at_index" btree (updated_at)
71
+ "is_pd12m_url_index" UNIQUE, btree (url)
72
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
73
+ residual_quantization = true
74
+ [build.internal]
75
+ lists = [20000]
76
+ build_threads = 6
77
+ spherical_centroids = false
78
+ ')
79
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
80
+ ```
81
+
82
+ ## Uses
83
+
84
+ This dataset is available for a wide range of applications.
85
+
86
+ ### Use with Chipmunk
87
+
88
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
89
+
90
+ ### Use directly in PostgreSQL
91
+
92
+ 1. Create Table
93
+
94
+ ```sql
95
+ CREATE TABLE IF NOT EXISTS is_pd12m (
96
+ id BIGSERIAL PRIMARY KEY,
97
+ url VARCHAR NOT NULL,
98
+ hash VARCHAR NOT NULL DEFAULT '',
99
+ caption VARCHAR NOT NULL DEFAULT '',
100
+ caption_long VARCHAR NOT NULL DEFAULT '',
101
+ origin_hash VARCHAR NOT NULL DEFAULT '',
102
+ origin_width BIGINT NOT NULL DEFAULT 0,
103
+ origin_height BIGINT NOT NULL DEFAULT 0,
104
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
105
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
106
+ processed_width BIGINT NOT NULL DEFAULT 0,
107
+ processed_height BIGINT NOT NULL DEFAULT 0,
108
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
109
+ exif JSONB NOT NULL DEFAULT '{}',
110
+ meta JSONB NOT NULL DEFAULT '{}',
111
+ source JSONB NOT NULL DEFAULT '[]',
112
+ vector halfvec(1152) DEFAULT NULL,
113
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
114
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
115
+ );
116
+ ```
117
+
118
+ 2. Load csv file to database
119
+
120
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
121
+
122
+ ```sql
123
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
124
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
125
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
126
+ ...
127
+ ```
128
+
129
+ 2.2. Load the dataset from the PostgreSQL server's file system:
130
+
131
+ ```sql
132
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
133
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
134
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
135
+ ...
136
+ ```
137
+
138
+ 3. Create Indexes
139
+
140
+ You need to create the following indexes for the best performance.
141
+
142
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
143
+
144
+ ```sql
145
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
156
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
157
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
158
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
159
+ residual_quantization = true
160
+ [build.internal]
161
+ lists = [20000]
162
+ build_threads = 6
163
+ spherical_centroids = false
164
+ $$);
165
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
166
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
167
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
168
+ ```
.history/README_20250513112617.md ADDED
@@ -0,0 +1,168 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - caption_long: A long caption for the image.
31
+ - width: The width of the image in pixels.
32
+ - height: The height of the image in pixels.
33
+ - mime_type: The MIME type of the image file.
34
+ - hash: The MD5 hash of the image file.
35
+ - license: The URL of the image license.
36
+ - source : The source organization of the image.
37
+
38
+ ```
39
+ \d is_pd12m;
40
+ Table "public.is_pd12m"
41
+ Column | Type | Collation | Nullable | Default
42
+ origin_width | bigint | | not null |
43
+ origin_height | bigint | | not null |
44
+ origin_storage_id | character varying(1024) | | not null |
45
+ processed_storage_id | character varying(1024) | | not null |
46
+ processed_width | bigint | | not null |
47
+ processed_height | bigint | | not null |
48
+ aspect_ratio | double precision | | not null |
49
+ exif | jsonb | | not null |
50
+ meta | jsonb | | not null |
51
+ created_at | timestamp without time zone | | not null |
52
+ updated_at | timestamp without time zone | | not null |
53
+ source | jsonb | | not null |
54
+ vector | halfvec(1152) | | | NULL::halfvec
55
+ Indexes:
56
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
57
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
58
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
59
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
60
+ "is_pd12m_created_at_index" btree (created_at)
61
+ "is_pd12m_exif_index" gin (exif)
62
+ "is_pd12m_meta_index" gin (meta)
63
+ "is_pd12m_origin_height_index" btree (origin_height)
64
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
65
+ "is_pd12m_origin_width_index" btree (origin_width)
66
+ "is_pd12m_processed_height_index" btree (processed_height)
67
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
68
+ "is_pd12m_processed_width_index" btree (processed_width)
69
+ "is_pd12m_source_index" gin (source)
70
+ "is_pd12m_updated_at_index" btree (updated_at)
71
+ "is_pd12m_url_index" UNIQUE, btree (url)
72
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
73
+ residual_quantization = true
74
+ [build.internal]
75
+ lists = [20000]
76
+ build_threads = 6
77
+ spherical_centroids = false
78
+ ')
79
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
80
+ ```
81
+
82
+ ## Uses
83
+
84
+ This dataset is available for a wide range of applications.
85
+
86
+ ### Use with Chipmunk
87
+
88
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
89
+
90
+ ### Use directly in PostgreSQL
91
+
92
+ 1. Create Table
93
+
94
+ ```sql
95
+ CREATE TABLE IF NOT EXISTS is_pd12m (
96
+ id BIGSERIAL PRIMARY KEY,
97
+ url VARCHAR NOT NULL,
98
+ hash VARCHAR NOT NULL DEFAULT '',
99
+ caption VARCHAR NOT NULL DEFAULT '',
100
+ caption_long VARCHAR NOT NULL DEFAULT '',
101
+ origin_hash VARCHAR NOT NULL DEFAULT '',
102
+ origin_width BIGINT NOT NULL DEFAULT 0,
103
+ origin_height BIGINT NOT NULL DEFAULT 0,
104
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
105
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
106
+ processed_width BIGINT NOT NULL DEFAULT 0,
107
+ processed_height BIGINT NOT NULL DEFAULT 0,
108
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
109
+ exif JSONB NOT NULL DEFAULT '{}',
110
+ meta JSONB NOT NULL DEFAULT '{}',
111
+ source JSONB NOT NULL DEFAULT '[]',
112
+ vector halfvec(1152) DEFAULT NULL,
113
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
114
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
115
+ );
116
+ ```
117
+
118
+ 2. Load csv file to database
119
+
120
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
121
+
122
+ ```sql
123
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
124
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
125
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
126
+ ...
127
+ ```
128
+
129
+ 2.2. Load the dataset from the PostgreSQL server's file system:
130
+
131
+ ```sql
132
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
133
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
134
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
135
+ ...
136
+ ```
137
+
138
+ 3. Create Indexes
139
+
140
+ You need to create the following indexes for the best performance.
141
+
142
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
143
+
144
+ ```sql
145
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
156
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
157
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
158
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
159
+ residual_quantization = true
160
+ [build.internal]
161
+ lists = [20000]
162
+ build_threads = 6
163
+ spherical_centroids = false
164
+ $$);
165
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
166
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
167
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
168
+ ```
.history/README_20250513112619.md ADDED
@@ -0,0 +1,170 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - caption_long: A long caption for the image.
31
+ - width: The width of the image in pixels.
32
+ - height: The height of the image in pixels.
33
+ - mime_type: The MIME type of the image file.
34
+ - hash: The MD5 hash of the image file.
35
+ - license: The URL of the image license.
36
+ - source : The source organization of the image.
37
+
38
+ ```
39
+ \d is_pd12m;
40
+ Table "public.is_pd12m"
41
+ Column | Type | Collation | Nullable | Default
42
+ ----------------------+-----------------------------+-----------+----------+---------------
43
+ caption_long | character varying | | not null |
44
+ origin_width | bigint | | not null |
45
+ origin_height | bigint | | not null |
46
+ origin_storage_id | character varying(1024) | | not null |
47
+ processed_storage_id | character varying(1024) | | not null |
48
+ processed_width | bigint | | not null |
49
+ processed_height | bigint | | not null |
50
+ aspect_ratio | double precision | | not null |
51
+ exif | jsonb | | not null |
52
+ meta | jsonb | | not null |
53
+ created_at | timestamp without time zone | | not null |
54
+ updated_at | timestamp without time zone | | not null |
55
+ source | jsonb | | not null |
56
+ vector | halfvec(1152) | | | NULL::halfvec
57
+ Indexes:
58
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
59
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
60
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
61
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
62
+ "is_pd12m_created_at_index" btree (created_at)
63
+ "is_pd12m_exif_index" gin (exif)
64
+ "is_pd12m_meta_index" gin (meta)
65
+ "is_pd12m_origin_height_index" btree (origin_height)
66
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
67
+ "is_pd12m_origin_width_index" btree (origin_width)
68
+ "is_pd12m_processed_height_index" btree (processed_height)
69
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
70
+ "is_pd12m_processed_width_index" btree (processed_width)
71
+ "is_pd12m_source_index" gin (source)
72
+ "is_pd12m_updated_at_index" btree (updated_at)
73
+ "is_pd12m_url_index" UNIQUE, btree (url)
74
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
75
+ residual_quantization = true
76
+ [build.internal]
77
+ lists = [20000]
78
+ build_threads = 6
79
+ spherical_centroids = false
80
+ ')
81
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
82
+ ```
83
+
84
+ ## Uses
85
+
86
+ This dataset is available for a wide range of applications.
87
+
88
+ ### Use with Chipmunk
89
+
90
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
91
+
92
+ ### Use directly in PostgreSQL
93
+
94
+ 1. Create Table
95
+
96
+ ```sql
97
+ CREATE TABLE IF NOT EXISTS is_pd12m (
98
+ id BIGSERIAL PRIMARY KEY,
99
+ url VARCHAR NOT NULL,
100
+ hash VARCHAR NOT NULL DEFAULT '',
101
+ caption VARCHAR NOT NULL DEFAULT '',
102
+ caption_long VARCHAR NOT NULL DEFAULT '',
103
+ origin_hash VARCHAR NOT NULL DEFAULT '',
104
+ origin_width BIGINT NOT NULL DEFAULT 0,
105
+ origin_height BIGINT NOT NULL DEFAULT 0,
106
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
107
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
108
+ processed_width BIGINT NOT NULL DEFAULT 0,
109
+ processed_height BIGINT NOT NULL DEFAULT 0,
110
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
111
+ exif JSONB NOT NULL DEFAULT '{}',
112
+ meta JSONB NOT NULL DEFAULT '{}',
113
+ source JSONB NOT NULL DEFAULT '[]',
114
+ vector halfvec(1152) DEFAULT NULL,
115
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
116
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
117
+ );
118
+ ```
119
+
120
+ 2. Load csv file to database
121
+
122
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
123
+
124
+ ```sql
125
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
126
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
127
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
128
+ ...
129
+ ```
130
+
131
+ 2.2. Load the dataset from the PostgreSQL server's file system:
132
+
133
+ ```sql
134
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
135
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
136
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
137
+ ...
138
+ ```
139
+
140
+ 3. Create Indexes
141
+
142
+ You need to create the following indexes for the best performance.
143
+
144
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
145
+
146
+ ```sql
147
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
156
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
157
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
158
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
160
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
161
+ residual_quantization = true
162
+ [build.internal]
163
+ lists = [20000]
164
+ build_threads = 6
165
+ spherical_centroids = false
166
+ $$);
167
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
168
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
169
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
170
+ ```
.history/README_20250513112620.md ADDED
@@ -0,0 +1,168 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - caption_long: A long caption for the image.
31
+ - width: The width of the image in pixels.
32
+ - height: The height of the image in pixels.
33
+ - mime_type: The MIME type of the image file.
34
+ - hash: The MD5 hash of the image file.
35
+ - license: The URL of the image license.
36
+ - source : The source organization of the image.
37
+
38
+ ```
39
+ \d is_pd12m;
40
+ Table "public.is_pd12m"
41
+ Column | Type | Collation | Nullable | Default
42
+ origin_width | bigint | | not null |
43
+ origin_height | bigint | | not null |
44
+ origin_storage_id | character varying(1024) | | not null |
45
+ processed_storage_id | character varying(1024) | | not null |
46
+ processed_width | bigint | | not null |
47
+ processed_height | bigint | | not null |
48
+ aspect_ratio | double precision | | not null |
49
+ exif | jsonb | | not null |
50
+ meta | jsonb | | not null |
51
+ created_at | timestamp without time zone | | not null |
52
+ updated_at | timestamp without time zone | | not null |
53
+ source | jsonb | | not null |
54
+ vector | halfvec(1152) | | | NULL::halfvec
55
+ Indexes:
56
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
57
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
58
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
59
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
60
+ "is_pd12m_created_at_index" btree (created_at)
61
+ "is_pd12m_exif_index" gin (exif)
62
+ "is_pd12m_meta_index" gin (meta)
63
+ "is_pd12m_origin_height_index" btree (origin_height)
64
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
65
+ "is_pd12m_origin_width_index" btree (origin_width)
66
+ "is_pd12m_processed_height_index" btree (processed_height)
67
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
68
+ "is_pd12m_processed_width_index" btree (processed_width)
69
+ "is_pd12m_source_index" gin (source)
70
+ "is_pd12m_updated_at_index" btree (updated_at)
71
+ "is_pd12m_url_index" UNIQUE, btree (url)
72
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
73
+ residual_quantization = true
74
+ [build.internal]
75
+ lists = [20000]
76
+ build_threads = 6
77
+ spherical_centroids = false
78
+ ')
79
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
80
+ ```
81
+
82
+ ## Uses
83
+
84
+ This dataset is available for a wide range of applications.
85
+
86
+ ### Use with Chipmunk
87
+
88
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
89
+
90
+ ### Use directly in PostgreSQL
91
+
92
+ 1. Create Table
93
+
94
+ ```sql
95
+ CREATE TABLE IF NOT EXISTS is_pd12m (
96
+ id BIGSERIAL PRIMARY KEY,
97
+ url VARCHAR NOT NULL,
98
+ hash VARCHAR NOT NULL DEFAULT '',
99
+ caption VARCHAR NOT NULL DEFAULT '',
100
+ caption_long VARCHAR NOT NULL DEFAULT '',
101
+ origin_hash VARCHAR NOT NULL DEFAULT '',
102
+ origin_width BIGINT NOT NULL DEFAULT 0,
103
+ origin_height BIGINT NOT NULL DEFAULT 0,
104
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
105
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
106
+ processed_width BIGINT NOT NULL DEFAULT 0,
107
+ processed_height BIGINT NOT NULL DEFAULT 0,
108
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
109
+ exif JSONB NOT NULL DEFAULT '{}',
110
+ meta JSONB NOT NULL DEFAULT '{}',
111
+ source JSONB NOT NULL DEFAULT '[]',
112
+ vector halfvec(1152) DEFAULT NULL,
113
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
114
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
115
+ );
116
+ ```
117
+
118
+ 2. Load csv file to database
119
+
120
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
121
+
122
+ ```sql
123
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
124
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
125
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
126
+ ...
127
+ ```
128
+
129
+ 2.2. Load the dataset from the PostgreSQL server's file system:
130
+
131
+ ```sql
132
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
133
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
134
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
135
+ ...
136
+ ```
137
+
138
+ 3. Create Indexes
139
+
140
+ You need to create the following indexes for the best performance.
141
+
142
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
143
+
144
+ ```sql
145
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
156
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
157
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
158
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
159
+ residual_quantization = true
160
+ [build.internal]
161
+ lists = [20000]
162
+ build_threads = 6
163
+ spherical_centroids = false
164
+ $$);
165
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
166
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
167
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
168
+ ```
.history/README_20250513112719.md ADDED
@@ -0,0 +1,165 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - caption_long: A long caption for the image.
31
+ - origin_hash: The MD5 hash of the original image file.
32
+ - origin_width: The width of the original image in pixels.
33
+ - origin_height: The height of the original image in pixels.
34
+
35
+ ```
36
+ \d is_pd12m;
37
+ Table "public.is_pd12m"
38
+ Column | Type | Collation | Nullable | Default
39
+ origin_width | bigint | | not null |
40
+ origin_height | bigint | | not null |
41
+ origin_storage_id | character varying(1024) | | not null |
42
+ processed_storage_id | character varying(1024) | | not null |
43
+ processed_width | bigint | | not null |
44
+ processed_height | bigint | | not null |
45
+ aspect_ratio | double precision | | not null |
46
+ exif | jsonb | | not null |
47
+ meta | jsonb | | not null |
48
+ created_at | timestamp without time zone | | not null |
49
+ updated_at | timestamp without time zone | | not null |
50
+ source | jsonb | | not null |
51
+ vector | halfvec(1152) | | | NULL::halfvec
52
+ Indexes:
53
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
54
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
55
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
56
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
57
+ "is_pd12m_created_at_index" btree (created_at)
58
+ "is_pd12m_exif_index" gin (exif)
59
+ "is_pd12m_meta_index" gin (meta)
60
+ "is_pd12m_origin_height_index" btree (origin_height)
61
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
62
+ "is_pd12m_origin_width_index" btree (origin_width)
63
+ "is_pd12m_processed_height_index" btree (processed_height)
64
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
65
+ "is_pd12m_processed_width_index" btree (processed_width)
66
+ "is_pd12m_source_index" gin (source)
67
+ "is_pd12m_updated_at_index" btree (updated_at)
68
+ "is_pd12m_url_index" UNIQUE, btree (url)
69
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
70
+ residual_quantization = true
71
+ [build.internal]
72
+ lists = [20000]
73
+ build_threads = 6
74
+ spherical_centroids = false
75
+ ')
76
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
77
+ ```
78
+
79
+ ## Uses
80
+
81
+ This dataset is available for a wide range of applications.
82
+
83
+ ### Use with Chipmunk
84
+
85
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
86
+
87
+ ### Use directly in PostgreSQL
88
+
89
+ 1. Create Table
90
+
91
+ ```sql
92
+ CREATE TABLE IF NOT EXISTS is_pd12m (
93
+ id BIGSERIAL PRIMARY KEY,
94
+ url VARCHAR NOT NULL,
95
+ hash VARCHAR NOT NULL DEFAULT '',
96
+ caption VARCHAR NOT NULL DEFAULT '',
97
+ caption_long VARCHAR NOT NULL DEFAULT '',
98
+ origin_hash VARCHAR NOT NULL DEFAULT '',
99
+ origin_width BIGINT NOT NULL DEFAULT 0,
100
+ origin_height BIGINT NOT NULL DEFAULT 0,
101
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
102
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
103
+ processed_width BIGINT NOT NULL DEFAULT 0,
104
+ processed_height BIGINT NOT NULL DEFAULT 0,
105
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
106
+ exif JSONB NOT NULL DEFAULT '{}',
107
+ meta JSONB NOT NULL DEFAULT '{}',
108
+ source JSONB NOT NULL DEFAULT '[]',
109
+ vector halfvec(1152) DEFAULT NULL,
110
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
111
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
112
+ );
113
+ ```
114
+
115
+ 2. Load csv file to database
116
+
117
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
118
+
119
+ ```sql
120
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
121
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
122
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
123
+ ...
124
+ ```
125
+
126
+ 2.2. Load the dataset from the PostgreSQL server's file system:
127
+
128
+ ```sql
129
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
130
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
131
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
132
+ ...
133
+ ```
134
+
135
+ 3. Create Indexes
136
+
137
+ You need to create the following indexes for the best performance.
138
+
139
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
140
+
141
+ ```sql
142
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
143
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
144
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
145
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
146
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
147
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
156
+ residual_quantization = true
157
+ [build.internal]
158
+ lists = [20000]
159
+ build_threads = 6
160
+ spherical_centroids = false
161
+ $$);
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
164
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
165
+ ```
.history/README_20250513112725.md ADDED
@@ -0,0 +1,171 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - caption_long: A long caption for the image.
31
+ - origin_hash: The MD5 hash of the original image file.
32
+ - origin_width: The width of the original image in pixels.
33
+ - origin_height: The height of the original image in pixels.
34
+ - origin_storage_id: The storage ID of the original image file.
35
+ - processed_storage_id: The storage ID of the processed image file.
36
+ - processed_width: The width of the processed image in pixels.
37
+ - processed_height: The height of the processed image in pixels.
38
+ - aspect_ratio: The aspect ratio of the image.
39
+ - exif: The EXIF data of the image.
40
+
41
+ ```
42
+ \d is_pd12m;
43
+ Table "public.is_pd12m"
44
+ Column | Type | Collation | Nullable | Default
45
+ origin_width | bigint | | not null |
46
+ origin_height | bigint | | not null |
47
+ origin_storage_id | character varying(1024) | | not null |
48
+ processed_storage_id | character varying(1024) | | not null |
49
+ processed_width | bigint | | not null |
50
+ processed_height | bigint | | not null |
51
+ aspect_ratio | double precision | | not null |
52
+ exif | jsonb | | not null |
53
+ meta | jsonb | | not null |
54
+ created_at | timestamp without time zone | | not null |
55
+ updated_at | timestamp without time zone | | not null |
56
+ source | jsonb | | not null |
57
+ vector | halfvec(1152) | | | NULL::halfvec
58
+ Indexes:
59
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
60
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
61
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
62
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
63
+ "is_pd12m_created_at_index" btree (created_at)
64
+ "is_pd12m_exif_index" gin (exif)
65
+ "is_pd12m_meta_index" gin (meta)
66
+ "is_pd12m_origin_height_index" btree (origin_height)
67
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
68
+ "is_pd12m_origin_width_index" btree (origin_width)
69
+ "is_pd12m_processed_height_index" btree (processed_height)
70
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
71
+ "is_pd12m_processed_width_index" btree (processed_width)
72
+ "is_pd12m_source_index" gin (source)
73
+ "is_pd12m_updated_at_index" btree (updated_at)
74
+ "is_pd12m_url_index" UNIQUE, btree (url)
75
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
76
+ residual_quantization = true
77
+ [build.internal]
78
+ lists = [20000]
79
+ build_threads = 6
80
+ spherical_centroids = false
81
+ ')
82
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
83
+ ```
84
+
85
+ ## Uses
86
+
87
+ This dataset is available for a wide range of applications.
88
+
89
+ ### Use with Chipmunk
90
+
91
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
92
+
93
+ ### Use directly in PostgreSQL
94
+
95
+ 1. Create Table
96
+
97
+ ```sql
98
+ CREATE TABLE IF NOT EXISTS is_pd12m (
99
+ id BIGSERIAL PRIMARY KEY,
100
+ url VARCHAR NOT NULL,
101
+ hash VARCHAR NOT NULL DEFAULT '',
102
+ caption VARCHAR NOT NULL DEFAULT '',
103
+ caption_long VARCHAR NOT NULL DEFAULT '',
104
+ origin_hash VARCHAR NOT NULL DEFAULT '',
105
+ origin_width BIGINT NOT NULL DEFAULT 0,
106
+ origin_height BIGINT NOT NULL DEFAULT 0,
107
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
108
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
109
+ processed_width BIGINT NOT NULL DEFAULT 0,
110
+ processed_height BIGINT NOT NULL DEFAULT 0,
111
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
112
+ exif JSONB NOT NULL DEFAULT '{}',
113
+ meta JSONB NOT NULL DEFAULT '{}',
114
+ source JSONB NOT NULL DEFAULT '[]',
115
+ vector halfvec(1152) DEFAULT NULL,
116
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
117
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
118
+ );
119
+ ```
120
+
121
+ 2. Load csv file to database
122
+
123
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
124
+
125
+ ```sql
126
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
127
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
128
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
129
+ ...
130
+ ```
131
+
132
+ 2.2. Load the dataset from the PostgreSQL server's file system:
133
+
134
+ ```sql
135
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
136
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
137
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
138
+ ...
139
+ ```
140
+
141
+ 3. Create Indexes
142
+
143
+ You need to create the following indexes for the best performance.
144
+
145
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
146
+
147
+ ```sql
148
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
156
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
157
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
158
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
160
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
162
+ residual_quantization = true
163
+ [build.internal]
164
+ lists = [20000]
165
+ build_threads = 6
166
+ spherical_centroids = false
167
+ $$);
168
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
169
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
170
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
171
+ ```
.history/README_20250513112727.md ADDED
@@ -0,0 +1,170 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - caption_long: A long caption for the image.
31
+ - origin_hash: The MD5 hash of the original image file.
32
+ - origin_width: The width of the original image in pixels.
33
+ - origin_height: The height of the original image in pixels.
34
+ - origin_storage_id: The storage ID of the original image file.
35
+ - processed_storage_id: The storage ID of the processed image file.
36
+ - processed_width: The width of the processed image in pixels.
37
+ - processed_height: The height of the processed image in pixels.
38
+ - aspect_ratio: The aspect ratio of the image.
39
+ - exif: The EXIF data of the image.
40
+ ```
41
+ \d is_pd12m;
42
+ Table "public.is_pd12m"
43
+ Column | Type | Collation | Nullable | Default
44
+ origin_width | bigint | | not null |
45
+ origin_height | bigint | | not null |
46
+ origin_storage_id | character varying(1024) | | not null |
47
+ processed_storage_id | character varying(1024) | | not null |
48
+ processed_width | bigint | | not null |
49
+ processed_height | bigint | | not null |
50
+ aspect_ratio | double precision | | not null |
51
+ exif | jsonb | | not null |
52
+ meta | jsonb | | not null |
53
+ created_at | timestamp without time zone | | not null |
54
+ updated_at | timestamp without time zone | | not null |
55
+ source | jsonb | | not null |
56
+ vector | halfvec(1152) | | | NULL::halfvec
57
+ Indexes:
58
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
59
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
60
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
61
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
62
+ "is_pd12m_created_at_index" btree (created_at)
63
+ "is_pd12m_exif_index" gin (exif)
64
+ "is_pd12m_meta_index" gin (meta)
65
+ "is_pd12m_origin_height_index" btree (origin_height)
66
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
67
+ "is_pd12m_origin_width_index" btree (origin_width)
68
+ "is_pd12m_processed_height_index" btree (processed_height)
69
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
70
+ "is_pd12m_processed_width_index" btree (processed_width)
71
+ "is_pd12m_source_index" gin (source)
72
+ "is_pd12m_updated_at_index" btree (updated_at)
73
+ "is_pd12m_url_index" UNIQUE, btree (url)
74
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
75
+ residual_quantization = true
76
+ [build.internal]
77
+ lists = [20000]
78
+ build_threads = 6
79
+ spherical_centroids = false
80
+ ')
81
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
82
+ ```
83
+
84
+ ## Uses
85
+
86
+ This dataset is available for a wide range of applications.
87
+
88
+ ### Use with Chipmunk
89
+
90
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
91
+
92
+ ### Use directly in PostgreSQL
93
+
94
+ 1. Create Table
95
+
96
+ ```sql
97
+ CREATE TABLE IF NOT EXISTS is_pd12m (
98
+ id BIGSERIAL PRIMARY KEY,
99
+ url VARCHAR NOT NULL,
100
+ hash VARCHAR NOT NULL DEFAULT '',
101
+ caption VARCHAR NOT NULL DEFAULT '',
102
+ caption_long VARCHAR NOT NULL DEFAULT '',
103
+ origin_hash VARCHAR NOT NULL DEFAULT '',
104
+ origin_width BIGINT NOT NULL DEFAULT 0,
105
+ origin_height BIGINT NOT NULL DEFAULT 0,
106
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
107
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
108
+ processed_width BIGINT NOT NULL DEFAULT 0,
109
+ processed_height BIGINT NOT NULL DEFAULT 0,
110
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
111
+ exif JSONB NOT NULL DEFAULT '{}',
112
+ meta JSONB NOT NULL DEFAULT '{}',
113
+ source JSONB NOT NULL DEFAULT '[]',
114
+ vector halfvec(1152) DEFAULT NULL,
115
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
116
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
117
+ );
118
+ ```
119
+
120
+ 2. Load csv file to database
121
+
122
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
123
+
124
+ ```sql
125
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
126
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
127
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
128
+ ...
129
+ ```
130
+
131
+ 2.2. Load the dataset from the PostgreSQL server's file system:
132
+
133
+ ```sql
134
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
135
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
136
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
137
+ ...
138
+ ```
139
+
140
+ 3. Create Indexes
141
+
142
+ You need to create the following indexes for the best performance.
143
+
144
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
145
+
146
+ ```sql
147
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
148
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
149
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
150
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
151
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
152
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
153
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
156
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
157
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
158
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
160
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
161
+ residual_quantization = true
162
+ [build.internal]
163
+ lists = [20000]
164
+ build_threads = 6
165
+ spherical_centroids = false
166
+ $$);
167
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
168
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
169
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
170
+ ```
.history/README_20250513112728.md ADDED
@@ -0,0 +1,176 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - caption_long: A long caption for the image.
31
+ - origin_hash: The MD5 hash of the original image file.
32
+ - origin_width: The width of the original image in pixels.
33
+ - origin_height: The height of the original image in pixels.
34
+ - origin_storage_id: The storage ID of the original image file.
35
+ - processed_storage_id: The storage ID of the processed image file.
36
+ - processed_width: The width of the processed image in pixels.
37
+ - processed_height: The height of the processed image in pixels.
38
+ - aspect_ratio: The aspect ratio of the image.
39
+ - exif: The EXIF data of the image.
40
+ - meta: The metadata of the image.
41
+ - source: The source organization of the image.
42
+ - vector: The vector embedding of the image.
43
+ - created_at: The creation time of the image.
44
+ - updated_at: The update time of the image.
45
+
46
+ ```
47
+ \d is_pd12m;
48
+ Table "public.is_pd12m"
49
+ Column | Type | Collation | Nullable | Default
50
+ origin_width | bigint | | not null |
51
+ origin_height | bigint | | not null |
52
+ origin_storage_id | character varying(1024) | | not null |
53
+ processed_storage_id | character varying(1024) | | not null |
54
+ processed_width | bigint | | not null |
55
+ processed_height | bigint | | not null |
56
+ aspect_ratio | double precision | | not null |
57
+ exif | jsonb | | not null |
58
+ meta | jsonb | | not null |
59
+ created_at | timestamp without time zone | | not null |
60
+ updated_at | timestamp without time zone | | not null |
61
+ source | jsonb | | not null |
62
+ vector | halfvec(1152) | | | NULL::halfvec
63
+ Indexes:
64
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
65
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
66
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
67
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
68
+ "is_pd12m_created_at_index" btree (created_at)
69
+ "is_pd12m_exif_index" gin (exif)
70
+ "is_pd12m_meta_index" gin (meta)
71
+ "is_pd12m_origin_height_index" btree (origin_height)
72
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
73
+ "is_pd12m_origin_width_index" btree (origin_width)
74
+ "is_pd12m_processed_height_index" btree (processed_height)
75
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
76
+ "is_pd12m_processed_width_index" btree (processed_width)
77
+ "is_pd12m_source_index" gin (source)
78
+ "is_pd12m_updated_at_index" btree (updated_at)
79
+ "is_pd12m_url_index" UNIQUE, btree (url)
80
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
81
+ residual_quantization = true
82
+ [build.internal]
83
+ lists = [20000]
84
+ build_threads = 6
85
+ spherical_centroids = false
86
+ ')
87
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
88
+ ```
89
+
90
+ ## Uses
91
+
92
+ This dataset is available for a wide range of applications.
93
+
94
+ ### Use with Chipmunk
95
+
96
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
97
+
98
+ ### Use directly in PostgreSQL
99
+
100
+ 1. Create Table
101
+
102
+ ```sql
103
+ CREATE TABLE IF NOT EXISTS is_pd12m (
104
+ id BIGSERIAL PRIMARY KEY,
105
+ url VARCHAR NOT NULL,
106
+ hash VARCHAR NOT NULL DEFAULT '',
107
+ caption VARCHAR NOT NULL DEFAULT '',
108
+ caption_long VARCHAR NOT NULL DEFAULT '',
109
+ origin_hash VARCHAR NOT NULL DEFAULT '',
110
+ origin_width BIGINT NOT NULL DEFAULT 0,
111
+ origin_height BIGINT NOT NULL DEFAULT 0,
112
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
113
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
114
+ processed_width BIGINT NOT NULL DEFAULT 0,
115
+ processed_height BIGINT NOT NULL DEFAULT 0,
116
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
117
+ exif JSONB NOT NULL DEFAULT '{}',
118
+ meta JSONB NOT NULL DEFAULT '{}',
119
+ source JSONB NOT NULL DEFAULT '[]',
120
+ vector halfvec(1152) DEFAULT NULL,
121
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
122
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
123
+ );
124
+ ```
125
+
126
+ 2. Load csv file to database
127
+
128
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
129
+
130
+ ```sql
131
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
132
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
133
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
134
+ ...
135
+ ```
136
+
137
+ 2.2. Load the dataset from the PostgreSQL server's file system:
138
+
139
+ ```sql
140
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
141
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
142
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
143
+ ...
144
+ ```
145
+
146
+ 3. Create Indexes
147
+
148
+ You need to create the following indexes for the best performance.
149
+
150
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
151
+
152
+ ```sql
153
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
156
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
157
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
158
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
160
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
164
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
165
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
166
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
167
+ residual_quantization = true
168
+ [build.internal]
169
+ lists = [20000]
170
+ build_threads = 6
171
+ spherical_centroids = false
172
+ $$);
173
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
174
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
175
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
176
+ ```
.history/README_20250513112733.md ADDED
@@ -0,0 +1,176 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - caption_long: A long caption for the image.
31
+ - origin_hash: The MD5 hash of the original image file.
32
+ - origin_width: The width of the original image in pixels.
33
+ - origin_height: The height of the original image in pixels.
34
+ - origin_storage_id: The storage ID of the original image file.
35
+ - processed_storage_id: The storage ID of the processed image file.
36
+ - processed_width: The width of the processed image in pixels.
37
+ - processed_height: The height of the processed image in pixels.
38
+ - aspect_ratio: The aspect ratio of the image.
39
+ - exif: The EXIF data of the image.
40
+ - meta: The metadata of the image.
41
+ - source: The source organization of the image.
42
+ - vector: The vector embedding of the image.
43
+ - created_at: The creation time of the image.
44
+ - updated_at: The update time of the image.
45
+
46
+ ```
47
+ \d is_pd12m;
48
+ Table "public.is_pd12m"
49
+ Column | Type | Collation | Nullable | Default
50
+ origin_width | bigint | | not null |
51
+ origin_height | bigint | | not null |
52
+ origin_storage_id | character varying(1024) | | not null |
53
+ processed_storage_id | character varying(1024) | | not null |
54
+ processed_width | bigint | | not null |
55
+ processed_height | bigint | | not null |
56
+ aspect_ratio | double precision | | not null |
57
+ exif | jsonb | | not null |
58
+ meta | jsonb | | not null |
59
+ created_at | timestamp without time zone | | not null |
60
+ updated_at | timestamp without time zone | | not null |
61
+ source | jsonb | | not null |
62
+ vector | halfvec(1152) | | | NULL::halfvec
63
+ Indexes:
64
+ "id_pd12m_pkey" PRIMARY KEY, btree (id)
65
+ "is_pd12m_aspect_ratio_index" btree (aspect_ratio)
66
+ "is_pd12m_caption_index" btree (caption) WHERE caption::text = ''::text
67
+ "is_pd12m_caption_long_index" btree (caption_long) WHERE caption_long::text = ''::text
68
+ "is_pd12m_created_at_index" btree (created_at)
69
+ "is_pd12m_exif_index" gin (exif)
70
+ "is_pd12m_meta_index" gin (meta)
71
+ "is_pd12m_origin_height_index" btree (origin_height)
72
+ "is_pd12m_origin_storage_id_index" btree (origin_storage_id)
73
+ "is_pd12m_origin_width_index" btree (origin_width)
74
+ "is_pd12m_processed_height_index" btree (processed_height)
75
+ "is_pd12m_processed_storage_id_index" btree (processed_storage_id)
76
+ "is_pd12m_processed_width_index" btree (processed_width)
77
+ "is_pd12m_source_index" gin (source)
78
+ "is_pd12m_updated_at_index" btree (updated_at)
79
+ "is_pd12m_url_index" UNIQUE, btree (url)
80
+ "is_pd12m_vector_index" vchordrq (vector halfvec_l2_ops) WITH (options='
81
+ residual_quantization = true
82
+ [build.internal]
83
+ lists = [20000]
84
+ build_threads = 6
85
+ spherical_centroids = false
86
+ ')
87
+ "is_pd12m_vector_null_index" btree (vector) WHERE vector IS NULL
88
+ ```
89
+
90
+ ## Uses
91
+
92
+ This dataset is available for a wide range of applications.
93
+
94
+ ### Use with Chipmunk
95
+
96
+ Click this link to learn how to load the dataset to [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk).
97
+
98
+ ### Use directly in PostgreSQL
99
+
100
+ 1. Create Table
101
+
102
+ ```sql
103
+ CREATE TABLE IF NOT EXISTS is_pd12m (
104
+ id BIGSERIAL PRIMARY KEY,
105
+ url VARCHAR NOT NULL,
106
+ hash VARCHAR NOT NULL DEFAULT '',
107
+ caption VARCHAR NOT NULL DEFAULT '',
108
+ caption_long VARCHAR NOT NULL DEFAULT '',
109
+ origin_hash VARCHAR NOT NULL DEFAULT '',
110
+ origin_width BIGINT NOT NULL DEFAULT 0,
111
+ origin_height BIGINT NOT NULL DEFAULT 0,
112
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
113
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
114
+ processed_width BIGINT NOT NULL DEFAULT 0,
115
+ processed_height BIGINT NOT NULL DEFAULT 0,
116
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
117
+ exif JSONB NOT NULL DEFAULT '{}',
118
+ meta JSONB NOT NULL DEFAULT '{}',
119
+ source JSONB NOT NULL DEFAULT '[]',
120
+ vector halfvec(1152) DEFAULT NULL,
121
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
122
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
123
+ );
124
+ ```
125
+
126
+ 2. Load csv file to database
127
+
128
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
129
+
130
+ ```sql
131
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
132
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
133
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
134
+ ...
135
+ ```
136
+
137
+ 2.2. Load the dataset from the PostgreSQL server's file system:
138
+
139
+ ```sql
140
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
141
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
142
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
143
+ ...
144
+ ```
145
+
146
+ 3. Create Indexes
147
+
148
+ You need to create the following indexes for the best performance.
149
+
150
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
151
+
152
+ ```sql
153
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
154
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
155
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
156
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
157
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
158
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
159
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
160
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
161
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
162
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
163
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
164
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
165
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
166
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
167
+ residual_quantization = true
168
+ [build.internal]
169
+ lists = [20000]
170
+ build_threads = 6
171
+ spherical_centroids = false
172
+ $$);
173
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
174
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
175
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
176
+ ```
.history/README_20250513122557.md ADDED
@@ -0,0 +1,128 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - feature-extraction
5
+ language:
6
+ - en
7
+ size_categories:
8
+ - 1M<n<10M
9
+ ---
10
+
11
+ # Dataset Card for `PD12M`
12
+
13
+ This is a curated PD12M dataset for use with the [Chipmunk](https://github.com/Intelligent-Internet/Chipmunk) project.
14
+
15
+ ## Dataset Details
16
+
17
+ ### Dataset Description
18
+
19
+ This dataset comprises a curated [Public Domain 12M](https://source.plus/pd12m) image collection, refined by filtering for active image links. EXIF data was extracted, and images underwent preprocessing and feature extraction using [SigLIP 2](https://huggingface.co/papers/2502.14786). All vector embeddings are normalized 16-bit half-precision vectors optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord).
20
+
21
+ ### Dataset Sources
22
+
23
+ This dataset is derived and organized from [Spawning/PD12M](http://huggingface.co/datasets/Spawning/PD12M). The original license information for the image can be found in the corresponding entry of the original database.
24
+
25
+ ## Dataset Structure
26
+
27
+ - id: A unique identifier for the image.
28
+ - url: The URL of the image.
29
+ - caption: A caption for the image.
30
+ - caption_long: A long caption for the image.
31
+ - origin_hash: The MD5 hash of the original image file.
32
+ - origin_width: The width of the original image in pixels.
33
+ - origin_height: The height of the original image in pixels.
34
+ - origin_storage_id: The storage ID of the original image file.
35
+ - processed_storage_id: The storage ID of the processed image file.
36
+ - processed_width: The width of the processed image in pixels.
37
+ - processed_height: The height of the processed image in pixels.
38
+ - aspect_ratio: The aspect ratio of the image.
39
+ - exif: The EXIF data of the image.
40
+ - meta: The metadata of the image.
41
+ - source: The source organization of the image.
42
+ - vector: The vector embedding of the image.
43
+ - created_at: The creation time of the image.
44
+ - updated_at: The update time of the image.
45
+
46
+ ## Uses
47
+
48
+ This dataset is available for a wide range of applications.
49
+
50
+ ### Use directly in PostgreSQL
51
+
52
+ 1. Create Table
53
+
54
+ ```sql
55
+ CREATE TABLE IF NOT EXISTS is_pd12m (
56
+ id BIGSERIAL PRIMARY KEY,
57
+ url VARCHAR NOT NULL,
58
+ hash VARCHAR NOT NULL DEFAULT '',
59
+ caption VARCHAR NOT NULL DEFAULT '',
60
+ caption_long VARCHAR NOT NULL DEFAULT '',
61
+ origin_hash VARCHAR NOT NULL DEFAULT '',
62
+ origin_width BIGINT NOT NULL DEFAULT 0,
63
+ origin_height BIGINT NOT NULL DEFAULT 0,
64
+ origin_storage_id VARCHAR(1024) DEFAULT NULL,
65
+ processed_storage_id VARCHAR(1024) NOT NULL DEFAULT '',
66
+ processed_width BIGINT NOT NULL DEFAULT 0,
67
+ processed_height BIGINT NOT NULL DEFAULT 0,
68
+ aspect_ratio DOUBLE PRECISION NOT NULL DEFAULT 0,
69
+ exif JSONB NOT NULL DEFAULT '{}',
70
+ meta JSONB NOT NULL DEFAULT '{}',
71
+ source JSONB NOT NULL DEFAULT '[]',
72
+ vector halfvec(1152) DEFAULT NULL,
73
+ created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
74
+ updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
75
+ );
76
+ ```
77
+
78
+ 2. Load csv file to database
79
+
80
+ 2.1. Load the dataset from local file system to a remote PostgreSQL server:
81
+
82
+ ```sql
83
+ \copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
84
+ \copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
85
+ \copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
86
+ ...
87
+ ```
88
+
89
+ 2.2. Load the dataset from the PostgreSQL server's file system:
90
+
91
+ ```sql
92
+ copy is_pd12m FROM 'data/0000000.csv' CSV HEADER;
93
+ copy is_pd12m FROM 'data/0000001.csv' CSV HEADER;
94
+ copy is_pd12m FROM 'data/0000002.csv' CSV HEADER;
95
+ ...
96
+ ```
97
+
98
+ 3. Create Indexes
99
+
100
+ You need to create the following indexes for the best performance.
101
+
102
+ The `vector` column is a halfvec(1152) column, which is a 16-bit half-precision vector optimized for L2 indexing with [vectorchord](https://github.com/tensorchord/vectorchord). You can get more information about the vector index from the [vectorchord](https://docs.vectorchord.ai/vectorchord/usage/indexing.html) documentation.
103
+
104
+ ```sql
105
+ CREATE UNIQUE INDEX IF NOT EXISTS is_pd12m_url_index ON is_pd12m (url);
106
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_width_index ON is_pd12m (origin_width);
107
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_height_index ON is_pd12m (origin_height);
108
+ CREATE INDEX IF NOT EXISTS is_pd12m_origin_storage_id_index ON is_pd12m (origin_storage_id);
109
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_storage_id_index ON is_pd12m (processed_storage_id);
110
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_width_index ON is_pd12m (processed_width);
111
+ CREATE INDEX IF NOT EXISTS is_pd12m_processed_height_index ON is_pd12m (processed_height);
112
+ CREATE INDEX IF NOT EXISTS is_pd12m_aspect_ratio_index ON is_pd12m (aspect_ratio);
113
+ CREATE INDEX IF NOT EXISTS is_pd12m_exif_index ON is_pd12m USING gin(exif);
114
+ CREATE INDEX IF NOT EXISTS is_pd12m_meta_index ON is_pd12m USING gin(meta);
115
+ CREATE INDEX IF NOT EXISTS is_pd12m_source_index ON is_pd12m USING gin(source);
116
+ CREATE INDEX IF NOT EXISTS is_pd12m_created_at_index ON is_pd12m (created_at);
117
+ CREATE INDEX IF NOT EXISTS is_pd12m_updated_at_index ON is_pd12m (updated_at);
118
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_index ON is_pd12m USING vchordrq (vector halfvec_l2_ops) WITH (options = $$
119
+ residual_quantization = true
120
+ [build.internal]
121
+ lists = [20000]
122
+ build_threads = 6
123
+ spherical_centroids = false
124
+ $$);
125
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_index ON is_pd12m (caption) WHERE caption = '';
126
+ CREATE INDEX IF NOT EXISTS is_pd12m_caption_long_index ON is_pd12m (caption_long) WHERE caption_long = '';
127
+ CREATE INDEX IF NOT EXISTS is_pd12m_vector_null_index ON is_pd12m (vector) WHERE vector IS NULL;
128
+ ```