BoxingWeb

This is the dataset repository for the paper BoxMind: Closed-loop AI strategy optimization for elite boxing validated in the 2024 Olympics.

Dataset Preparation

Download BoxingWeb Dataset

Download the BoxingWeb dataset from Tsinghua Cloud.

Extract Dataset

Extract the downloaded dataset and place it in the data/ directory with the following structure:

data/
└── boxingweb/
    ├── data_train/
    │   └── 202.../
    │       ├── 202....mp4
    │       ├── 202....pkl
    │       └── video_event.json
    └── data_test/
        └── ...

Installation (Boxing Punch Classification)

Requirements

• Python >= 3.7
• PyTorch >= 2.0

Clone the Repository

git clone https://github.com/gouba2333/BoxingWeb
cd BoxingWeb

Model Preparation

Download I3D Pretrained Weights

Download the I3D pretrained weights from here.

Place the downloaded weights in the checkpoint/ directory:

checkpoint/
└── rgb_imagenet.pt

Training

To train the model, run:

python train.py

Citation

If you find this work useful, please cite our paper:

@article{wang2026boxmind,
  title={BoxMind: Closed-loop AI strategy optimization for elite boxing validated in the 2024 Olympics},
  author={Kaiwen Wang, Kaili Zheng, Rongrong Deng, Qingmin Fan, Milin Zhang, Zongrui Li, Xuesi Zhou, Bo Han, Liren Chen, Chenyi Guo, Ji Wu},
  journal={arXiv preprint arXiv:2601.11492},
  year={2026},
  url={http://arxiv.org/abs/2601.11492}
}

License

Please refer to the LICENSE file for more information.