Research map / Video / Video segmentation and tracking
Accelerating the Training of Video Super-resolution Models
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 |
|---|---|---|---|
| Use hierarchical spatial and temporal cycles with large minibatch training | BasicVSR-M · REDS4 | speedup: 6.2× | Helped |
| Use large minibatch, spatial cycle, and temporal cycle together | BasicVSR-M · REDS4 | PSNR: 30.91 -> 30.90 | No clear effect |
| Use hierarchical spatial and temporal cycles with large minibatch training | BasicVSR · REDS4 | speedup: 3.1× | Helped |
| Use multigrid and large minibatch training | EDVR-M · REDS4 | speedup: 2.3× | Helped |
| Use multigrid and large minibatch training | BasicVSR · Vimeo-90K | PSNR: 37.18 -> 37.32 | Helped |
| Use multigrid and large minibatch training | EDVR-M · Vimeo-90K | PSNR: 35.10 -> 35.22 | Helped |
| Add linear learning-rate scaling for minibatch 64 (ablation) | BasicVSR-M · REDS4 | PSNR: 30.64 -> 30.93 | Helped |
| Add warmup to large-minibatch training (ablation) | BasicVSR-M · REDS4 | PSNR: 30.81 -> 30.93 | Helped |
| Use constant rather than no warmup (ablation) | BasicVSR-M · REDS4 | PSNR: 30.81 -> 30.91 | Helped |
| Increase minibatch from 16 to 32 (ablation) | BasicVSR-M · REDS4 | speedup: 2.0× | Helped |
| Increase minibatch from 16 to 48 (ablation) | BasicVSR-M · REDS4 | speedup: 2.9× | Helped |
| Use more spatial stages, 32/48/64 rather than 32/64 (ablation) | BasicVSR-M Multi-S · REDS4 | PSNR: 30.90 -> 30.89 | Hurt |
Closest papers by meaning
- A Multigrid Method for Efficiently Training Video Models (CVPR 2020 · similarity 0.90)
- SAVSR: Arbitrary-Scale Video Super-Resolution via a Learned Scale-Adaptive Network (AAAI 2024 · similarity 0.90)
- Time Without Time: Pseudo-Temporal Representation for Space-Time Super-Resolution (CVPR 2026 · similarity 0.89)
- VideoINR: Learning Video Implicit Neural Representation for Continuous Space-Time Super-Resolution (CVPR 2022 · similarity 0.89)
- DOVE: Efficient One-Step Diffusion Model for Real-World Video Super-Resolution (NeurIPS 2025 · similarity 0.88)
- Sequence Matters: Harnessing Video Models in 3D Super-Resolution (AAAI 2025 · similarity 0.87)
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.