Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -37,9 +37,12 @@ dataset_info:
|
|
| 37 |
---
|
| 38 |
|
| 39 |
# Dataset Card for aRefCOCO
|
| 40 |
-
[](https://creativecommons.org/licenses/by/4.0/)
|
| 44 |
|
| 45 |
**aRefCOCO** (Ambiguous RefCOCO) is a dataset specifically constructed for Referring Image Segmentation (RIS), focusing on **referential ambiguity** that frequently arises in real-world application. It introduces **object-distracting expressions**, which involve multiple entities with contextual cues, and **category-implicit expressions**, where the object class is not explicitly stated. Each entity is paired with an image, a target segmentation mask, multiple referring descriptions, and supporting metadata such as bounding boxes and category labels. In addition to the original benchmark used for evaluation, aRefCOCO now provides an extended **train split** to support model training and further research on referential ambiguity in referring segmentation and related tasks.
|
|
@@ -99,7 +102,7 @@ print(f"Descriptions: {sample['descriptions']}")
|
|
| 99 |
|
| 100 |
This Hugging Face repository contains the dataset in **Parquet/Arrow format** for easy loading.
|
| 101 |
|
| 102 |
-
For alternative formats and implementations, please visit the **[GitHub Repository](https://github.com/zhenjiemao/
|
| 103 |
- Custom PyTorch Dataset class (`refdataset/refdataset.py`)
|
| 104 |
- Source images and masks in original quality
|
| 105 |
- JSONL metadata files
|
|
|
|
| 37 |
---
|
| 38 |
|
| 39 |
# Dataset Card for aRefCOCO
|
| 40 |
+
[](https://zhenjiemao.github.io/SaFiRe/)
|
| 41 |
+
[](https://arxiv.org/pdf/2510.10160)
|
| 42 |
+
[](https://arxiv.org/abs/2510.10160)
|
| 43 |
+
[](https://github.com/zhenjiemao/SaFiRe)
|
| 44 |
+
[](https://github.com/zhenjiemao/aRefCOCO)
|
| 45 |
+
[](https://huggingface.co/datasets/zhenjiemao/aRefCOCO)
|
| 46 |
[](https://creativecommons.org/licenses/by/4.0/)
|
| 47 |
|
| 48 |
**aRefCOCO** (Ambiguous RefCOCO) is a dataset specifically constructed for Referring Image Segmentation (RIS), focusing on **referential ambiguity** that frequently arises in real-world application. It introduces **object-distracting expressions**, which involve multiple entities with contextual cues, and **category-implicit expressions**, where the object class is not explicitly stated. Each entity is paired with an image, a target segmentation mask, multiple referring descriptions, and supporting metadata such as bounding boxes and category labels. In addition to the original benchmark used for evaluation, aRefCOCO now provides an extended **train split** to support model training and further research on referential ambiguity in referring segmentation and related tasks.
|
|
|
|
| 102 |
|
| 103 |
This Hugging Face repository contains the dataset in **Parquet/Arrow format** for easy loading.
|
| 104 |
|
| 105 |
+
For alternative formats and implementations, please visit the **[GitHub Repository](https://github.com/zhenjiemao/aRefCOCO)** which includes:
|
| 106 |
- Custom PyTorch Dataset class (`refdataset/refdataset.py`)
|
| 107 |
- Source images and masks in original quality
|
| 108 |
- JSONL metadata files
|