Research map / Video / Video segmentation and tracking
Unsupervised Flow-Aligned Sequence-to-Sequence Learning for Video Restoration
Jing Lin, Xiaowan Hu, Yuanhao Cai, Haoqian Wang, Youliang Yan, Xueyi Zou, Yulun Zhang, Luc Van Gool
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 |
|---|---|---|---|
| Use S2SVR instead of the strongest listed baseline, IconVSR | Video super-resolution · REDS4 | PSNR: 31.67 -> 31.96 | Improved |
| Use S2SVR instead of VSR-Transformer | Video super-resolution · Vimeo-90K-T | PSNR: 37.71 -> 37.63 | Hurt |
| Use S2SVR instead of TSP | Video deblurring · GOPRO | PSNR: 31.67 -> 31.81 | Improved |
| Use S2SVR instead of TSP | Video deblurring · GOPRO | SSIM: 0.9279 -> 0.9231 | Hurt |
| Use S2SVR instead of STDF-R3L | Compressed video enhancement · Test sequences at QP=37 | ΔPSNR (dB): 0.830 -> 0.925 | Improved |
| Use S2SVR instead of STDF-R3L | Compressed video enhancement · Test sequences at QP=37 | ΔSSIM: 151 -> 176 | Improved |
| Train Pwclite with unsupervised distillation rather than FlyingChairs supervision (ablation) | S2SVR flow estimator · REDS4 | PSNR (dB): 31.79 -> 31.96 | Improved |
| Replace supervised RAFT with the unsupervised-distillation Pwclite flow estimator (ablation) | S2SVR flow estimator · REDS4 | PSNR (dB): 31.88 -> 31.96 | Improved |
| Use S2SVR rather than EDVR as sequence length increases | Video restoration on REDS4 · Sequences of length 5, 15, 30, and 50 | PSNR: 28.05 -> 28.10; 28.80 -> 29.25; 28.01 -> 28.53; 27.77 -> 28.34 | Improved |
Closest papers by meaning
- Temporal-Consistent Video Restoration with Pre-trained Diffusion Models (AAAI 2026 · similarity 0.86)
- FlowVid: Taming Imperfect Optical Flows for Consistent Video-to-Video Synthesis (CVPR 2024 · similarity 0.85)
- One-Step Diffusion for Detail-Rich and Temporally Consistent Video Super-Resolution (NeurIPS 2025 · similarity 0.84)
- Taming generative video models for zero-shot optical flow extraction (NeurIPS 2025 · similarity 0.84)
- FMA-Net: Flow-Guided Dynamic Filtering and Iterative Feature Refinement with Multi-Attention for Joint Video Super-Resolution and Deblurring (CVPR 2024 · similarity 0.84)
- Scaling Video-Language Models to 10K Frames via Hierarchical Differential Distillation (ICML 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.
