Research map / Video / Video segmentation and tracking
DeLiVR: Differential Spatiotemporal Lie Bias for Efficient Video Deraining
Shuning Sun, Jialang Lu, Xiang Chen, Jichao Wang, Dianjie Lu, Guijuan Zhang, Guangwei Gao, Zhuoran Zheng
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add spatial Lie relative bias to plain spatiotemporal Transformer (ablation) | DeLiVR baseline · NTURain | PSNR: 29.21 -> 32.58 | Improved |
| add differential temporal bias (group displacement) on top of spatial bias (ablation) | DeLiVR · NTURain | PSNR: 32.58 -> 33.14 | Improved |
| add temporal decay and banded attention mask (ablation) | DeLiVR · NTURain | PSNR: 33.14 -> 34.06 | Improved |
| replace RAFT optical-flow bias maps with SO(2) Lie rotation bias (ablation) | 12-layer Transformer backbone · NTURain | PSNR: +2.43 dB | Improved |
| use Lie-group rotation-prediction model instead of flow-guided recurrent models | DeLiVR · WeatherBench (real-world) | PSNR: 24.10 -> 26.56 | Improved |
| DeLiVR vs VDMamba on real-world data | DeLiVR · WeatherBench (real-world) | PSNR: 23.91 -> 26.56 | Improved |
| DeLiVR vs VDMamba on synthetic NTURain | DeLiVR · NTURain | PSNR: 36.29 -> 34.06 | Hurt |
| DeLiVR vs MFGAN on hard synthetic rain | DeLiVR · Rain-Syn-Complex | SSIM: 0.763 -> 0.733 | Hurt |
| Lie bias model vs best baseline on perceptual metric in real rain | DeLiVR · WeatherBench (real-world) | LPIPS: 0.334 -> 0.358 | Hurt |
| use DeLiVR rather than larger prior models | DeLiVR · NTURain | Params (M): 12.70 -> 2.64 (vs VDMamba) | Improved |
Closest papers by meaning
- Rethinking Video Rain Streak Removal: A New Synthesis Model and a Deraining Network with Video Rain Prior (ECCV 2022 · similarity 0.85)
- Bias for Action: Video Implicit Neural Representations with Bias Modulation (CVPR 2025 · similarity 0.84)
- VideoSPatS: Video SPatiotemporal Splines for Disentangled Occlusion, Appearance and Motion Modeling and Editing (CVPR 2025 · similarity 0.84)
- Semi-Supervised State-Space Model with Dynamic Stacking Filter for Real-World Video Deraining (CVPR 2025 · similarity 0.84)
- StreamFlow: Streamlined Multi-Frame Optical Flow Estimation for Video Sequences (NeurIPS 2024 · similarity 0.83)
- Orientation-anchored Hyper-Gaussian for 4D Reconstruction from Casual Videos (NeurIPS 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
