1 Show Lab, National University of Singapore
2 National University of Singapore
| A New Dataset | A New Baseline |
UENR-600K is a 600,000-frame paired dataset for nighttime video deraining, rendered in Unreal Engine 5.3 with physically grounded 3D rain particle simulation. We benchmark seven existing deraining methods retrained on UENR-600K, and introduce a video diffusion baseline adapted from Wan 2.2 with LoRA finetuning.
@article{UENR600K,
title={UENR-600K: A Large-Scale Physically Grounded Dataset for Nighttime Video Deraining},
author={Pei Yang and Hai Ci and Beibei Lin and Yiren Song and Mike Zheng Shou},
year={2026},
eprint={2604.04402},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2604.04402},
}