Research map / Video / Video segmentation and tracking
Understanding Deformable Alignment in Video Super-Resolution
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 offset-fidelity loss to constrain offsets using optical flow (ablation) | EDVR-L · REDS4 | PSNR: 28.753 -> 30.480; +1.727 | Helped |
| Add offset-fidelity loss to constrain offsets using optical flow (ablation) | EDVR-L · Vimeo-90K-T | PSNR: 33.632 -> 35.223; +1.591 | Helped |
| Warp features rather than images (ablation) | Flow-based alignment · REDS4 | PSNR: 29.155 -> 29.992; +0.837 | Helped |
| Increase TDAN offset count from N=1 to N=9 (ablation) | TDAN · Vimeo-90K-T | PSNR: 33.313 -> 33.483 | Helped |
| Increase TDAN offset count from N=9 to N=25 (ablation) | TDAN · Vimeo-90K-T | PSNR: 33.483 -> 33.540 | Helped |
| Increase flow-based alignment offset count from N=1 to N=2 (ablation) | Flow-based alignment · Vimeo-90K-T | PSNR: 32.835 -> 32.973 | Helped |
| Increase flow-based alignment offset count from N=2 to N=3 (ablation) | Flow-based alignment · Vimeo-90K-T | PSNR: 32.973 -> 33.017 | Helped |
| Use the warping-plus-convolution decomposition instead of DCN (ablation) | EDVR-M, G=1, k=1 · REDS4 | PSNR: 29.979 -> 29.992; Difference +0.013 | No clear effect |
| Use the warping-plus-convolution decomposition instead of DCN (ablation) | EDVR-M, G=1, k=3 · REDS4 | PSNR: 30.199 -> 30.240; Difference +0.041 | No clear effect |
| Use the warping-plus-convolution decomposition instead of DCN (ablation) | EDVR-M, G=8, k=1 · REDS4 | PSNR: 30.183 -> 30.179; Difference -0.004 | No clear effect |
| Use the warping-plus-convolution decomposition instead of DCN (ablation) | EDVR-M, G=8, k=3 · REDS4 | PSNR: 30.264 -> 30.231; Difference -0.033 | No clear effect |
| Increase deformable-group offset diversity (ablation) | EDVR-M · REDS4 | PSNR: up to 0.19 dB | Helped |
Closest papers by meaning
- TDAN: Temporally-Deformable Alignment Network for Video Super-Resolution (CVPR 2020 · similarity 0.89)
- Enhancing Video Super-Resolution via Implicit Resampling-based Alignment (CVPR 2024 · similarity 0.86)
- Rethinking Alignment in Video Super-Resolution Transformers (NeurIPS 2022 · similarity 0.86)
- Sequence Matters: Harnessing Video Models in 3D Super-Resolution (AAAI 2025 · similarity 0.84)
- RivuletMLP: An MLP-based Architecture for Efficient Compressed Video Quality Enhancement (CVPR 2025 · similarity 0.84)
- Restage4D: Reanimating Deformable 3D Reconstruction from a Single Video (NeurIPS 2025 · 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.