Research map / Video / Video segmentation and tracking
LaverNet: Lightweight All-in-one Video Restoration via Selective Propagation
Haiyu Zhao, Yiwen Shan, Wangmeng Zuo, Yuanbiao Gou, Xi Peng
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 the Selective Propagation Module (SPM) (ablation) | LaverNet · DAVIS-test (t=12) | PSNR: 32.32 -> 32.60 | Improved |
| add the Selective Propagation Module (SPM) (ablation) | LaverNet · DAVIS-test (t=12) | SSIM: 0.8896 -> 0.8967 | Improved |
| use degradation-guided attention block in LEM instead of dense block (ablation) | LaverNet · DAVIS-test (t=12) | PSNR: 32.23 -> 32.60 | Improved |
| use degradation-guided attention block in LEM instead of dense block (ablation) | LaverNet · DAVIS-test (t=12) | SSIM: 0.8911 -> 0.8967 | Improved |
| increase channel width from 32 to 48 (LaverNet-L) (ablation) | LaverNet · DAVIS-test (t=12) | PSNR: 32.60 -> 32.92 | Improved |
| decrease channel width from 32 to 16 (LaverNet-T) (ablation) | LaverNet · DAVIS-test (t=12) | PSNR: 32.60 -> 31.88 | Hurt |
| use LaverNet-T instead of EvrNet at comparable parameter count | EvrNet · DAVIS-test (t=12) | PSNR: 31.22 -> 31.88 | Improved |
| use LaverNet instead of PromptRestorer | PromptRestorer · DAVIS-test | PSNR: up to 1.2dB in PSNR and 0.0272 in SSIM | Improved |
| use LaverNet instead of EDVR | EDVR · DAVIS-test (noise & compression) | PSNR: 29.58 -> 34.13 | Improved |
| use LaverNet instead of WDiffusion | WDiffusion · Set8 (t=24) | PSNR: 30.82 -> 30.56 | Hurt |
| use LaverNet instead of WDiffusion | WDiffusion · Set8 (t=12) | PSNR: 30.02 -> 30.08 | Improved |
| use LaverNet instead of ViWS-Net | ViWS-Net · DAVIS-test (t=12) | PSNR: 16.48 -> 32.60 | Improved |
Closest papers by meaning
- Recurrent Video Restoration Transformer with Guided Deformable Attention (NeurIPS 2022 · similarity 0.87)
- SeedVR: Seeding Infinity in Diffusion Transformer Towards Generic Video Restoration (CVPR 2025 · similarity 0.85)
- AverNet: All-in-one Video Restoration for Time-varying Unknown Degradations (NeurIPS 2024 · similarity 0.85)
- One-Step Diffusion for Detail-Rich and Temporally Consistent Video Super-Resolution (NeurIPS 2025 · similarity 0.85)
- FMA-Net: Flow-Guided Dynamic Filtering and Iterative Feature Refinement with Multi-Attention for Joint Video Super-Resolution and Deblurring (CVPR 2024 · similarity 0.85)
- Upscale-A-Video: Temporal-Consistent Diffusion Model for Real-World Video Super-Resolution (CVPR 2024 · similarity 0.84)
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.
