BestWJH/VRU_Basketball

This is a basketball dataset from AVS-VRU, which contains three dynamic basketball court scenes collected in Guangzhou (gz), Dongguan (dg), and Long360. It features large-scale motion, making it well-suited for evaluating dynamic scene reconstruction methods.

Each scene contains 36 viewpoints, with videos captured at 1080p or 4K resolution for each viewpoint. The Guangzhou (gz) and Dongguan (dg) sequences are approximately 10 seconds long, each consisting of 250 frames at 1080p. The Long360 sequence contains 700 frames at 4K resolution. Viewpoints 0, 10, 20, and 30 are used for testing, while the remaining viewpoints are used for training.

It was presented in the paper Swift4D and LocalDyGS.

If you find it useful, we would appreciate it if you could cite our paper:

@inproceedings{
wu2025swiftd,
title={Swift4D: Adaptive divide-and-conquer Gaussian Splatting for compact and efficient reconstruction of dynamic scene},
author={Jiahao Wu and Rui Peng and Zhiyan Wang and Lu Xiao and Luyang Tang and Jinbo Yan and Kaiqiang Xiong and Ronggang Wang},
booktitle={The Thirteenth International Conference on Learning Representations},
year={2025},
url={https://openreview.net/forum?id=c1RhJVTPwT}
}

@inproceedings{wu2025localdygs,
  title={LocalDyGS: Multi-view Global Dynamic Scene Modeling via Adaptive Local Implicit Feature Decoupling},
  author={Wu, Jiahao and Peng, Rui and Jiao, Jianbo and Yang, Jiayu and Tang, Luyang and Xiong, Kaiqiang and Liang, Jie and Yan, Jinbo and Liu, Runling and Wang, Ronggang},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={9519--9529},
  year={2025}
}