evan6007 commited on
Commit
d53674a
·
1 Parent(s): ebe33ec

Please enter the commit message for your changes. Lines starting

Browse files

with '#' will be ignored, and an empty message aborts the commit.

On branch main
Your branch is up to date with 'origin/main'.

Changes to be committed:
new file: README.md
new file: count_pcb_images.py

Files changed (2) hide show
  1. README.md +113 -0
  2. count_pcb_images.py +62 -0
README.md ADDED
@@ -0,0 +1,113 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ pretty_name: Innodisk PCB Image Dataset
3
+ license: cc-by-nc-4.0
4
+ tags:
5
+ - image-classification
6
+ - pcb
7
+ - defect-detection
8
+ - innodisk
9
+ - small-sample-learning
10
+ configs:
11
+ - name: default
12
+ data_files:
13
+ train:
14
+ - "train_20_20/*.jpg"
15
+ - "train_20_20_R/*.jpg"
16
+ - "train_40_20/*.jpg"
17
+ - "train_40_20_R/*.jpg"
18
+ - "train_60_20/*.jpg"
19
+ - "train_80_20/*.jpg"
20
+ - "train_100_20/*.jpg"
21
+ test:
22
+ - "test/*.jpg"
23
+ - "test_R/*.jpg"
24
+ - "test_canny/*.jpg"
25
+ - "test_R_canny/*.jpg"
26
+ val:
27
+ - "val/*.jpg"
28
+ - "val_R/*.jpg"
29
+ ---
30
+
31
+ # 🚀🔬 Innodisk PCB Image Dataset
32
+
33
+ The **Innodisk PCB Image Dataset** is a collaboration between **Innodisk Corporation** and Tamkang University.
34
+ Designed for research on **printed‑circuit‑board (PCB) defect inspection** and **small‑sample learning**. 🧑‍🔧💡
35
+
36
+ | 📂 Split | 🖼️ Images | 📁 Source directories |
37
+ |----------|-----------|-----------------------|
38
+ | 🏋️‍♂️ Train | **500** | 7 `train_*` dirs |
39
+ | 🧪 Test | **240** | 4 `test*` dirs |
40
+ | 📝 Val | **30** | 2 `val*` dirs |
41
+
42
+ 🎯 *Resolutions vary; the modal size is **188 × 128 px**.*
43
+
44
+ 🔖 Filenames embed pseudo‑labels: **`pass`** (=acceptable) or **`ng`** (=defective).
45
+
46
+ ---
47
+
48
+ ## 🗂️ Folder layout
49
+
50
+ ```
51
+ Innodisk_PCB_datasets/
52
+ ├── train_20_20/ # 🏋️‍♂️
53
+ ├── train_20_20_R/
54
+ ├── train_40_20/
55
+ ├── train_40_20_R/
56
+ ├── train_60_20/
57
+ ├── train_80_20/
58
+ ├── train_100_20/
59
+ ├── test/ # 🧪
60
+ ├── test_R/
61
+ ├── test_canny/
62
+ ├── test_R_canny/
63
+ ├── val/ # 📝
64
+ └── val_R/
65
+ ```
66
+
67
+ ---
68
+
69
+ ## ⚡ Quick start
70
+
71
+ ```python
72
+ from datasets import load_dataset
73
+
74
+ # 🔄 Load images
75
+ ds_train = load_dataset("evan6007/Innodisk_PCB_datasets", split="train")
76
+
77
+ # 🏷️ Derive labels from filenames
78
+ def add_label(example):
79
+ fname = example["image"].filename.lower()
80
+ example["label"] = 0 if "ng" in fname else 1 # 0 = defective, 1 = acceptable
81
+ return example
82
+
83
+ ds_train = ds_train.map(add_label)
84
+ ```
85
+
86
+ ---
87
+
88
+ ## 🏭 Image sources & preprocessing
89
+
90
+ 1. 📷 **AOI line cameras** – raw production images (`train_*`, `test*`, `val*`).
91
+ 2. ✂️ **ROI versions** (`*_R`) – cropped component regions to reduce background noise.
92
+ 3. 🖊️ **Canny edge maps** (`*_canny`) – generated with `cv2.Canny(img, 50, 150)`.
93
+
94
+ ---
95
+
96
+ ## 📜 License
97
+
98
+ Released under **Creative Commons Attribution‑NonCommercial 4.0** (CC BY‑NC‑4.0).
99
+ Commercial applications require prior approval from Innodisk Corporation.
100
+
101
+ ---
102
+
103
+ ## 📚 Citation
104
+
105
+ ```bibtex
106
+ @dataset{innodisk_pcb_2025,
107
+ author = {Innodisk Corporation and Hsiao, Chao-Hsiang},
108
+ title = {Innodisk PCB Image Dataset},
109
+ year = {2025},
110
+ url = {https://huggingface.co/datasets/evan6007/Innodisk_PCB_datasets},
111
+ note = {CC BY-NC 4.0}
112
+ }
113
+ ```
count_pcb_images.py ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ """
3
+ Count images and summarize resolutions for Innodisk PCB dataset.
4
+
5
+ Usage:
6
+ python count_pcb_images.py /path/to/Innodisk_PCB_datasets
7
+ """
8
+
9
+ import sys
10
+ from pathlib import Path
11
+ from collections import defaultdict
12
+ from PIL import Image
13
+
14
+ IMAGE_EXTS = {'.jpg', '.jpeg', '.png', '.bmp', '.tif', '.tiff'}
15
+
16
+ def scan_dataset(root: Path):
17
+ splits = defaultdict(list) # split_name -> list[Path]
18
+ for p in root.glob('*'):
19
+ if not p.is_dir():
20
+ continue
21
+ name = p.name.lower()
22
+ if name.startswith('train'):
23
+ split = 'train'
24
+ elif name.startswith('test'):
25
+ split = 'test'
26
+ elif name.startswith('val'):
27
+ split = 'val'
28
+ else:
29
+ # unknown folder, skip
30
+ continue
31
+ for img in p.rglob('*'):
32
+ if img.suffix.lower() in IMAGE_EXTS:
33
+ splits[split].append(img)
34
+ return splits
35
+
36
+ def summarize(splits):
37
+ print("\n===== Image Counts =====")
38
+ for split, files in splits.items():
39
+ print(f"{split.capitalize():5s}: {len(files):,} images")
40
+ print("\n===== Resolution Stats =====")
41
+ for split, files in splits.items():
42
+ res_counter = defaultdict(int)
43
+ for img_path in files:
44
+ try:
45
+ with Image.open(img_path) as im:
46
+ res_counter[im.size] += 1
47
+ except Exception:
48
+ continue
49
+ print(f"\n[{split.capitalize()}]")
50
+ for (w, h), cnt in sorted(res_counter.items(), key=lambda x: (-x[1], x[0])):
51
+ print(f"{w}×{h}: {cnt:,}")
52
+
53
+ if __name__ == "__main__":
54
+ if len(sys.argv) != 2:
55
+ print("Usage: python count_pcb_images.py /path/to/Innodisk_PCB_datasets")
56
+ sys.exit(1)
57
+ root = Path(sys.argv[1])
58
+ if not root.exists():
59
+ print("Path does not exist:", root)
60
+ sys.exit(1)
61
+ splits = scan_dataset(root)
62
+ summarize(splits)