Datasets:

iLearn-Lab
/

Finebadminton-20K

pretty_name: Finebadminton-20K
language:
  - en
annotations_creators:
  - expert-generated
  - machine-generated
multilinguality:
  - monolingual
license: apache-2.0
configs:
  - config_name: annotations
    default: true
    data_files:
      - split: train
        path: finebadminton-20K/*.json

Finebadminton-20K

Finebadminton-20K is a public subset of the FineBadminton dataset for fine-grained badminton video understanding.

This release contains:

70 full-match videos
2,066 annotated rallies
20,757 annotated hit instances

The released 20K subset corresponds to 20,757 of the 33,325 hit annotations available in the full FineBadminton collection.

Dataset Structure

videos/: the 70 full-length source videos included in this release.
finebadminton-20K/: annotation JSON files, one file per source video.
benchmark_source_videos.txt: the list of source videos from this release that were used to construct FineBadmintonBenchmark.

Relation to FineBadmintonBenchmark

Part of the previously released FineBadmintonBenchmark was constructed from videos contained in this dataset release.

In total, 20 source videos in Finebadminton-20K were used for FineBadmintonBenchmark. The full list is provided in benchmark_source_videos.txt.

Citation

@inproceedings{he2025finebadminton,
  title={Finebadminton: A multi-level dataset for fine-grained badminton video understanding},
  author={He, Xusheng and Liu, Wei and Ma, Shanshan and Liu, Qian and Ma, Chenghao and Wu, Jianlong},
  booktitle={Proceedings of the 33rd ACM International Conference on Multimedia},
  pages={12776--12783},
  year={2025}
}

@misc{he2025finebadminton_arxiv,
  title={Finebadminton: A multi-level dataset for fine-grained badminton video understanding},
  author={He, Xusheng and Liu, Wei and Ma, Shanshan and Liu, Qian and Ma, Chenghao and Wu, Jianlong},
  year={2025},
  eprint={2508.07554},
  archivePrefix={arXiv},
  primaryClass={cs.MM},
  url={https://arxiv.org/abs/2508.07554}
}