Research map / 3D vision / Depth and stereo
SUNet: Symmetric Undistortion Network for Rolling Shutter Correction
Bin Fan, Yuchao Dai, Mingyi He
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 context-aware cost volume (search range 4) (ablation) | SUNet · Carla-RS with mask (CRM) | PSNR: 21.87 -> 29.28 | Improved |
| add context-aware cost volume (search range 4) (ablation) | SUNet · Fastec-RS | PSNR: 25.21 -> 28.34 | Improved |
| increase cost volume search range from 4 to 6 (ablation) | SUNet · Fastec-RS | PSNR: 28.34 -> 28.14 | No clear effect |
| use a 7x7 kernel at first pyramid level instead of removing the layer (ablation) | SUNet · Fastec-RS | PSNR: 27.83 -> 28.34 | Improved |
| multiply undistortion flows by time offsets (ablation) | SUNet · Carla-RS with mask (CRM) | PSNR: 29.28 -> 28.56 | Hurt |
| add ground-truth-supervised consistency loss (ablation) | SUNet · Fastec-RS | PSNR: 27.89 -> 28.34 | Improved |
| replace self-supervised consistency with GT-supervised consistency (ablation) | SUNet · Fastec-RS | PSNR: 28.02 -> 28.34 | Improved |
| add reconstruction loss L_r (ablation) | SUNet · Carla-RS with mask (CRM) | PSNR: 28.00 -> 29.28 | Improved |
| add perceptual loss L_p (ablation) | SUNet · Fastec-RS | PSNR: 28.20 -> 28.34 | Improved |
| add smoothness loss L_s (ablation) | SUNet · Carla-RS without mask (CR) | PSNR: 28.07 -> 29.18 | Improved |
| use SUNet symmetric two-frame design instead of DSUN | SUNet vs DSUN · Fastec-RS | PSNR: 26.52 -> 28.34 | Improved |
Closest papers by meaning
- Deep Shutter Unrolling Network (CVPR 2020 · similarity 0.91)
- Rolling Shutter Inversion: Bring Rolling Shutter Images to High Framerate Global Shutter Video (TPAMI 2022 · similarity 0.89)
- Learning Structure-And-Motion-Aware Rolling Shutter Correction (CVPR 2019 · similarity 0.88)
- Joint Appearance and Motion Learning for Efficient Rolling Shutter Correction (CVPR 2023 · similarity 0.88)
- 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.
