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

```bibtex
@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}
}
```