Research map / Video / Video segmentation and tracking
Joint Appearance and Motion Learning for Efficient Rolling Shutter Correction
Bin Fan, Yuxin Mao, Yuchao Dai, Zhexiong Wan, Qi Liu
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 |
|---|---|---|---|
| Replace the two-stage RSC pipeline with single-stage joint appearance and motion learning | JAMNet · BS-RSC | PSNR: 28.23 -> 32.93 | Improved |
| Add a synthesis-based branch to the joint decoder (ablation) | JAMNet · Fastec-RS | PSNR: 28.34 -> 28.70 | Improved |
| Add a warping-based branch to the joint decoder (ablation) | JAMNet · Fastec-RS | PSNR: 26.10 -> 28.70 | Improved |
| Add reconstruction loss Lr (ablation) | JAMNet · Fastec-RS | PSNR: 26.36 -> 28.70 | Improved |
| Add perceptual loss Lp (ablation) | JAMNet · Fastec-RS | PSNR: 28.45 -> 28.70 | Improved |
| Add multi-scale consistency loss Lmc (ablation) | JAMNet · Fastec-RS | PSNR: 27.84 -> 28.70 | Improved |
| Add total variation loss Ltv (ablation) | JAMNet · Fastec-RS | PSNR: 28.57 -> 28.70 | Improved |
| Use a 5-level rather than 3-level feature pyramid (ablation) | JAMNet · Fastec-RS | PSNR: 27.28 -> 28.70 | Improved |
| Add transformer-based motion embedding instead of no motion embedding (ablation) | JAMNet · Fastec-RS | PSNR: 28.02 -> 28.70 | Improved |
| Replace cost-volume motion embedding with transformer motion embedding (ablation) | JAMNet · Fastec-RS | PSNR: 28.31 -> 28.70 | Improved |
| Pass a hidden state between pyramid levels (ablation) | JAMNet · Fastec-RS | PSNR: 28.01 -> 28.70 | Improved |
| Apply vertical flip + inverse order augmentation (ablation) | JAMNet · Fastec-RS | PSNR: 28.33 -> 28.70 | Improved |
Closest papers by meaning
- Learning Structure-And-Motion-Aware Rolling Shutter Correction (CVPR 2019 · similarity 0.91)
- Deep Shutter Unrolling Network (CVPR 2020 · similarity 0.90)
- Towards Rolling Shutter Correction and Deblurring in Dynamic Scenes (CVPR 2021 · similarity 0.90)
- Learning Adaptive Warping for Real-World Rolling Shutter Correction (CVPR 2022 · similarity 0.89)
- Rolling Shutter Correction with Intermediate Distortion Flow Estimation (CVPR 2024 · similarity 0.87)
- Motion Blur Decomposition with Cross-shutter Guidance (CVPR 2024 · similarity 0.85)
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.
