Research map / Video / Video segmentation and tracking
Video Enhancement with Task-Oriented Flow
Tianfan Xue, Baian Chen, Jiajun Wu, Donglai Wei, William T. Freeman
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 |
|---|---|---|---|
| joint training of flow with the task network (TOFlow vs Fixed Flow) (ablation) | TOFlow · Vimeo interpolation | PSNR: 29.09 -> 33.53 | Improved |
| joint training of flow with the task network (ablation) | TOFlow · Vimeo denoising (Mixed) | PSNR: 31.85 -> 33.51 | Improved |
| joint training of flow with the task network (ablation) | TOFlow · Vimeo denoising (Gauss15) | PSNR: 36.25 -> 36.63 | Improved |
| joint training of flow with the task network (ablation) | TOFlow · Vimeo deblocking (q=60) | PSNR: 33.06 -> 34.02 | Improved |
| joint training of flow with the task network (ablation) | TOFlow · Vimeo-SR (7 frames) | PSNR: 31.81 -> 33.08 | Improved |
| add occlusion mask network (ablation) | TOFlow · Vimeo interpolation | PSNR: 33.53 -> 33.73 | Improved |
| add occlusion mask network (ablation) | Fixed Flow · Vimeo interpolation | PSNR: 29.09 -> 30.10 | Improved |
| use flow network trained on deblocking instead of denoising (ablation) | TOFlow denoising · Vimeo denoising (Gauss25) | PSNR: 34.89 -> 25.74 | Hurt |
| add more input frames (3 to 5 to 7) (ablation) | TOFlow · Vimeo-SR | PSNR: 32.66 -> 33.04 -> 33.08 | Improved |
| change downsampling kernel from cubic to box to Gaussian (ablation) | TOFlow · Vimeo-SR | PSNR: 33.08 -> 32.08 -> 31.15 | Hurt |
| task-oriented fine-tuning of flow network (TOFlow vs Fixed Flow vs EpicFlow) (ablation) | flow estimation · Sintel flow | EPE (all): 14.120 -> 16.638 (EpicFlow 4.115) | Hurt |
| replace SpyNet with FlowNetC as flow module, joint training (ablation) | TOFlow · Vimeo denoising | PSNR: 24.685 -> 24.689 | No clear effect |
Closest papers by meaning
- Offline and Online Optical Flow Enhancement for Deep Video Compression (AAAI 2024 · similarity 0.86)
- DPFlow: Adaptive Optical Flow Estimation with a Dual-Pyramid Framework (CVPR 2025 · similarity 0.85)
- FlowFeat: Pixel-Dense Embedding of Motion Profiles (NeurIPS 2025 · similarity 0.85)
- SAMFlow: Eliminating Any Fragmentation in Optical Flow with Segment Anything Model (AAAI 2024 · similarity 0.85)
- FlowVid: Taming Imperfect Optical Flows for Consistent Video-to-Video Synthesis (CVPR 2024 · similarity 0.85)
- Optical-Flow Guided Prompt Optimization for Coherent Video Generation (CVPR 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.
