Model Comparison

Comparing Deraining Methods

All models trained on UENR-600K, tested on real nighttime rain videos from the Internet.1 Drag any slider to compare — all sliders move together.

Dataset Comparison

Training Data Matters

Our baseline trained on different datasets, tested on Internet videos.1

References

  1. ^1 Test videos sourced from Pexels (free stock video).
  2. ^2 K. Zhang, D. Li, W. Luo, W. Ren, and W. Liu, "Enhanced Spatio-Temporal Interaction Learning for Video Deraining: Faster and Better," IEEE Trans. Pattern Anal. Mach. Intell. (TPAMI), 2022.
  3. ^3 S. Wang, L. Zhu, H. Fu, J. Qin, C.-B. Schönlieb, W. Feng, and S. Wang, "Rethinking Video Rain Streak Removal: A New Synthesis Model and a Deraining Network with Video Rain Prior," in Proc. ECCV, 2022.
  4. ^4 F. Zhang, S. You, Y. Li, and Y. Fu, "Learning Rain Location Prior for Nighttime Deraining," in Proc. IEEE/CVF ICCV, 2023.
  5. ^5 A. Ghasemabadi, M. K. Janjua, M. Salameh, and D. Niu, "Learning Truncated Causal History Model for Video Restoration," in Advances in NeurIPS, 2024.
  6. ^6 J. Zhuang, Y. Luo, X. Zhao, T. Jiang, and B. Guo, "UConNet: Unsupervised Controllable Network for Image and Video Deraining," in Proc. ACM MM, 2022.
  7. ^7 O. Özdenizci and R. Legenstein, "Restoring Vision in Adverse Weather Conditions with Patch-Based Denoising Diffusion Models," IEEE Trans. Pattern Anal. Mach. Intell. (TPAMI), 2023.
  8. ^8 B. Lin, Y. Jin, W. Yan, W. Ye, Y. Yuan, S. Zhang, and R. T. Tan, "NightRain: Nighttime Video Deraining via Adaptive-Rain-Removal and Adaptive-Correction," in Proc. AAAI, 2024.