Datasets:
Acknowledge license to accept the repository
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
Log in or Sign Up to review the conditions and access this dataset content.
Dataset Card for SportsHHI
Dataset Summary
SportsHHI is a dataset for video human-human interaction detection. Please refer to SportsHHI: A Dataset for Human-Human Interaction Detection in Sports Videos for more details. Please refer to this repository for training and evaluation.
Supported Tasks and Leaderboards
Human-Human Interaction Detection
Details about training and evaluation can be found in the GitHub Repository.
Languages
The class labels in the dataset are in English.
Dataset Structure
The dataset contains frames.zip and annotations.
frames.zip contains extracted video frames.
annotations contains annotaions about human boxes, interactions, and interaction classes.
If you have any network issues, please refer to pan.txt for data downloading.
Data Splits
| train | validation | |
|---|---|---|
| # of tubes | 38527 | 12122 |
Dataset Creation
Curation Rationale
comprehending high-level inter- actions between humans is crucial for understanding com- plex multi-person videos, such as sports and surveillance videos. To address this issue, authors propose a new video visual relation detection task: video human-human interaction detection, and build a dataset named SportsHHI for it.
Source Data
Initial Data Collection and Normalization
We carefully selected 80 basketball and 80 volleyball videos from the MultiSports dataset to cover various types of games including men’s, women’s, national team, and club games. The average length of the videos is 603 frames and the frame rate of the videos is 25FPS. All videos have a high resolution of 720P.
Annotations
Please refer to SportsHHI: A Dataset for Human-Human Interaction Detection in Sports Videos for more information.
Additional Information
Dataset Curators
Authors of SportsHHI: A Dataset for Human-Human Interaction Detection in Sports Videos
Tao Wu
Runyu He
Gangshan Wu
Limin Wang
Licensing Information
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Citation Information
If you find this dataset useful, please cite as
@misc{wu2024sportshhi,
title={SportsHHI: A Dataset for Human-Human Interaction Detection in Sports Videos},
author={Tao Wu and Runyu He and Gangshan Wu and Limin Wang},
year={2024},
eprint={2404.04565},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
- Downloads last month
- 44