AffordSplat / README.md
Weizm's picture
Update README.md
6bc6c76 verified
metadata
language:
  - en
size_categories:
  - 10K<n<100K
ArXiv:
  - https://arxiv.org/abs/2504.11218

In this repository, we present 3DAffordSplat, the first large-scale, multi-modal dataset tailored for 3DGS-based affordance reasoning. This dataset includes 23k Gaussian instances, 8k point cloud instances, and 6k manually annotated affordance labels, encompassing 21 object categories and 18 affordance types.

Dataset Structure

After downloading, the data structure should be as follows:

—Seen
    ├── train
    │   ├── bag
    │   │   ├── Gaussian
    │   │   │   └── GS_0017.ply
    │   │   │       ......
    │   │   ├── PointCloud
    │   │   │   └── PC_0001.ply
    │   │   │       ......
    │   │   ├── contain
    │   │   │   ├── GS_anno_0017.ply
    │   │   │   ├── PC_anno_0001.json
    │   │   │       ......
    │   │   └── grasp
    │   │       ......
    │   └── bed
    │       ......
    │
    ├── val
    │   ├── bag
    │   │   ├── Gaussian
    │   │   │   └── GS_0009.ply
    │   │   │       ......
    │   │   ├── contain
    │   │   │   └── GS_anno_0009.ply
    │   │   │       ......
    │   │   └── grasp
    │   │       ......
    │   └── bed
    │       ......
    │
    └── test
        ├── bag
        │   ├── Gaussian
        │   │   └── GS_0001.ply
        │   │       ......
        │   ├── contain
        │   │   └── GS_anno_0001.ply
        │   │       ......
        │   └── grasp
        │       ......
        └── bed
            ......

—Affordance-Question.csv
—obj_aff_structure.json
—UnSeen_test.json
—UnSeen_train.json

Dataset Details

For more information on detailed statistics and the methodology of AffordSplat, please refer to the following resources:

Additionally, we sincerely thank Guantian Liu, Yao Xiao, Xinyu Li, Kecheng Liang and Yipeng Ouyang for their contributions.

Contact

This project is for research purpose only, please contact us for the licence of commercial use. For any other questions please contact ([email protected], [email protected] or [email protected]).

Citation

If you use this data, please cite our paper.

@misc{wei20253daffordsplatefficientaffordancereasoning,
      title={3DAffordSplat: Efficient Affordance Reasoning with 3D Gaussians}, 
      author={Zeming wei and Junyi Lin and Yang Liu and Weixing Chen and Jingzhou Luo and Guanbin Li and Liang Lin},
      year={2025},
      eprint={2504.11218},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2504.11218}, 
}

Acknowledgement

The construction of AffordSplat dataset is based on 3DAffordanceNet, LASO and ShapeSplat. We sincerely thank them for their contributions to the community.