Research map / Label-efficient and robust learning / Recognition and long-tail
Occlusion Aware Unsupervised Learning of Optical Flow
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 occlusion-aware masking to the photometric loss (ablation) | Unsupervised FlowNetS optical flow · Flying Chairs and Sintel Final train | EPE: 5.11 to 4.51 on Flying-Chairs; 7.82 to 7.32 on MPI-Sintel Final | Helped |
| Add enlarged-neighbor search during warping (ablation) | Full unsupervised model · Flying Chairs and Sintel Final train | EPE: 3.76 to 3.30 on Flying-Chairs; 6.54 to 6.34 on MPI-Sintel Final | Helped |
| Add warped inputs and photometric-error inputs at decoder refinement stages (ablation) | Unsupervised FlowNetS optical flow · Flying Chairs and Sintel Final train | EPE: 5.11 to 4.62 on Flying-Chairs; 7.82 to 7.33 on MPI-Sintel Final | Helped |
| Apply contrast enhancement preprocessing (ablation) | Unsupervised FlowNetS optical flow · Flying Chairs and Sintel Final train | EPE: 4.14 to 3.30 on Flying-Chairs; 7.08 to 6.34 on MPI-Sintel Final | Helped |
| Fine-tune the Chairs-pretrained model on Sintel | Ours · Sintel Clean test | EPE: 8.02 -> 7.95 | No clear effect |
| Use 60,000 KITTI raw-video samples instead of 25,000 (ablation) | Ours-KITTI · KITTI | EPE | No clear effect |
| Use SpyNet network structure with the unsupervised training method (ablation) | Unsupervised optical flow model · KITTI | EPE | No clear effect |
| Combine occlusion handling, enlarged search, modified FlowNet, and contrast enhancement (ablation) | Unsupervised FlowNetS optical flow · Flying Chairs and Sintel | EPE: 5.11 -> 3.30 on Chairs; 6.93 -> 5.23 on Sintel Clean train; 7.82 -> 6.34 on Sintel Final train | Helped |
Closest papers by meaning
- UnFlow: Unsupervised Learning of Optical Flow With a Bidirectional Census Loss (AAAI 2018 · similarity 0.91)
- Unsupervised Deep Learning for Optical Flow Estimation (AAAI 2017 · similarity 0.91)
- OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation (CVPR 2024 · similarity 0.90)
- UnSAMFlow: Unsupervised Optical Flow Guided by Segment Anything Model (CVPR 2024 · similarity 0.90)
- Recurrent Partial Kernel Network for Efficient Optical Flow Estimation (AAAI 2024 · similarity 0.87)
- M2Flow: A Motion Information Fusion Framework for Enhanced Unsupervised Optical Flow Estimation in Autonomous Driving (AAAI 2025 · similarity 0.86)
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.