Repetition Action Counting Dataset

ShanghaiTech University

Github HomePage

[Download(OneDrive,extraction code:repcount)] [Download(BaiduNetDisk, extraction code: svip)]

Introduction

We introduce a novel repetition action counting dataset called RepCount that contains videos with significant variations in length and allows for multiple kinds of anomaly cases. These video data collaborate with fine-grained annotations that indicate the beginning and end of each action period. Furthermore, the dataset consists of two subsets namely Part-A and Part-B. The videos in Part-A are fetched from YouTube, while the others in Part-B record simulated physical examinations by junior school students and teachers.

Dataset Format

This dataset consists of two subsets namely Part-A and Part-B. For part-A, we collected 1041 videos from YouTube. The type of actionsincludes workout activities (squatting, pulling-up, frontraising, etc.), athletic events (rowing, pommel horse, etc.) and other repetitive actions (soccer juggling). For Part-B, we record the videos of exercise such as sitting- and pulling-up done by volunteers. It is constructed for the validation of model generalization. In brief, we provide 1451 videos collaborated with 19280 annotations. The videos from our dataset havea n average length of 39.359 seconds.


Data info:
  • The RepCount Part-Adataset divide into trian set ,valid setandtest set. The RepCount Part-Bdataset only contains a test set that is used to test generalization.
  • The videos information includes the URL of source video, action type and total number of frames.
  • The labels are mainly composed of count and location where count indicates the number of reptition action and location indicates the position of each cycle on the time axis.
    • You can use the open-source script YouTube download to to download source videos. In addition, you can use the official script to edit them to keep only useful clips.

    Citation

    If you find this useful, please cite our work as follows: 
    @article{hu2022transrac,
title={TransRAC: Encoding Multi-scale Temporal Correlation with Transformers for Repetitive Action Counting},
author={Hu, Huazhang and Dong, Sixun and Zhao, Yiqun and Lian, Dongze and Li, Zhengxin and Gao, Shenghua},
journal={arXiv preprint arXiv:2204.01018},
year={2022}
}