Research map / Label-efficient and robust learning / Self-supervised and contrastive
SelFlow: Self-Supervised Learning of Optical Flow
Check 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 |
|---|---|---|---|
| Distill reliable non-occluded flow to train predictions for newly occluded pixels using superpixel noise (ablation) | SelFlow · KITTI 2012 train, occluded pixels | EPE: 31.29 to 6.95 | Helped |
| Add occlusion handling (ablation) | SelFlow · KITTI 2012 ALL | EPE: 6.52 -> 3.32 | Helped |
| Use multiple frames as input (ablation) | SelFlow · Sintel Final ALL | EPE: 4.06 -> 3.87 | Helped |
| Use superpixel noise instead of rectangle noise for occlusion hallucination (ablation) | SelFlow · Sintel Final ALL | EPE: 3.99 -> 3.87 | Helped |
| Add unsupervised pre-training before supervised fine-tuning (ablation) | SelFlow+ft · Sintel Clean validation | EPE: 1.97 -> 1.50 | Helped |
| Add unsupervised pre-training before supervised fine-tuning (ablation) | SelFlow+ft · Sintel Final validation | EPE: 2.68 -> 2.41 | Helped |
| Add unsupervised pre-training before supervised fine-tuning (ablation) | SelFlow+ft · KITTI 2012 validation | EPE: 3.93 -> 1.55 | Helped |
| Add unsupervised pre-training before supervised fine-tuning (ablation) | SelFlow+ft · KITTI 2015 validation | EPE: 3.10 -> 1.86 | Helped |
| Use SelFlow instead of DDFlow+ft | Unsupervised optical flow · Sintel Final test | EPE: 7.40 -> 6.57 | Helped |
| Use SelFlow instead of DDFlow+ft | Unsupervised optical flow · KITTI 2015 test | Fl-all: 14.29% -> 14.19% | Helped |
| Use SelFlow+ft instead of ContinualFlow+ft | Supervised fine-tuned optical flow · Sintel Final test | EPE: 4.52 -> 4.26 | Helped |
| Use SelFlow+ft instead of PWC-Net+ft-axXiv | Supervised fine-tuned optical flow · KITTI 2012 test | Fl-all: 6.82% -> 6.19% | Mixed |
Closest papers by meaning
- Self-Supervised Monocular Scene Flow Estimation (CVPR 2020 · similarity 0.90)
- Flow2Stereo: Effective Self-Supervised Learning of Optical Flow and Stereo Matching (CVPR 2020 · similarity 0.90)
- Recurrent Partial Kernel Network for Efficient Optical Flow Estimation (AAAI 2024 · similarity 0.88)
- UnSAMFlow: Unsupervised Optical Flow Guided by Segment Anything Model (CVPR 2024 · similarity 0.87)
- OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation (CVPR 2024 · similarity 0.87)
- DPFlow: Adaptive Optical Flow Estimation with a Dual-Pyramid Framework (CVPR 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.