Abdulmalik Alquwayfili commited on
Configure Dataset viewer tabs
Browse files
README.md
CHANGED
|
@@ -1,3 +1,38 @@
|
|
| 1 |
-
---
|
| 2 |
-
|
| 3 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
dataset_info:
|
| 3 |
+
- config_name: coco
|
| 4 |
+
data_files:
|
| 5 |
+
- split: test
|
| 6 |
+
path: coco/mscoco_dense.jsonl
|
| 7 |
+
- config_name: flickr30k
|
| 8 |
+
data_files:
|
| 9 |
+
- split: test
|
| 10 |
+
path: flickr30k/flickr30k_dense.jsonl
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# LARE: Dense-Set
|
| 14 |
+
|
| 15 |
+
Dense-Set is a benchmark dataset tailored for text-to-image retrieval in densely crowded scenes. It provides hard, fine-grained, and low-attention subsets extracted from MS-COCO and Flickr30k.
|
| 16 |
+
|
| 17 |
+
This dataset corresponds to the **LARE** (Low-Attention Guided Dual Encoding) methodology.
|
| 18 |
+
|
| 19 |
+
## Subsets
|
| 20 |
+
|
| 21 |
+
You can explore the two subsets perfectly separated using the dropdown menu at the top of the Dataset Viewer!
|
| 22 |
+
|
| 23 |
+
- **`coco`**: 3,089 images with dense captions and rare-class mappings.
|
| 24 |
+
- **`flickr30k`**: 2,477 images with dense captions and rare-class mappings.
|
| 25 |
+
|
| 26 |
+
### Loading the Data
|
| 27 |
+
```python
|
| 28 |
+
from datasets import load_dataset
|
| 29 |
+
|
| 30 |
+
# Load COCO Dense-Set
|
| 31 |
+
coco_ds = load_dataset("AbdulmalekDS/Dense-Set", "coco")
|
| 32 |
+
|
| 33 |
+
# Load Flickr30k Dense-Set
|
| 34 |
+
flickr_ds = load_dataset("AbdulmalekDS/Dense-Set", "flickr30k")
|
| 35 |
+
```
|
| 36 |
+
|
| 37 |
+
## Images
|
| 38 |
+
The images referenced in this dataset (`filename` column) correspond to the standard **MS-COCO val2014** and **Flickr30k** image sets. Researchers evaluating on Dense-Set should download the standard image corpus from their original maintainers and utilize our JSON Lines metadata to evaluate LARE configurations.
|