VirtualCrowd
VirtualCrowd is released under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0) for non-commercial academic research and educational use. Access is granted after review by the Crowd4D and DyCrowd authors.
VirtualCrowd is provided for quantitative evaluation of crowd reconstruction methods. We do not provide a training split. The dataset contains four scenes, and each scene includes two camera views. The motions in the two views are not paired or temporally corresponding. For more details about the dataset construction and setting, please refer to the DyCrowd paper.
Data Use
By requesting access, you agree to use this dataset only for non-commercial academic research or educational purposes. Redistribution, commercial use, and re-identification attempts are prohibited.
Citation
If you use VirtualCrowd, please cite Crowd4D and DyCrowd:
@article{wen2025dycrowd,
author = {Hao Wen, Hongbo Kang, Jian Ma, Jing Huang, Yuanwang Yang, Haozhe Lin, Yu-Kun Lai and Kun Li},
title = {DyCrowd: Towards Dynamic Crowd Reconstruction from a Large-scene Video},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
year = {2025}
}
@inproceedings{kang2026crowd4d,
author = {Hongbo Kang, Tianyi Zhou, Qingyang Yang, Hongwei Wen, Jing Huang, Yu-Kun Lai and Kun Li},
title = {Crowd4D: Scene-Aware Monocular 4D Crowd Reconstruction},
booktitle = {International Conference on Machine Learning (ICML)},
year = {2026}
}
- Downloads last month
- 3