Research map / Video / Video segmentation and tracking
Learning To Estimate Hidden Motions With Global Motion Aggregation
Shihao Jiang, Dylan Campbell, Yao Lu, Hongdong Li, Richard Hartley
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 |
|---|---|---|---|
| add global motion aggregation (GMA) | RAFT · Sintel Clean train, occluded pixels | AEPE: 5.36 -> 4.25 | Improved |
| add GMA | RAFT · Sintel Clean train, out-of-frame occlusion | AEPE: 7.01 -> 5.03 | Improved |
| add GMA | RAFT · Sintel Clean train, non-occluded pixels | AEPE: 0.32 -> 0.29 | Improved |
| add GMA | RAFT · Sintel test Final | AEPE: 2.86 to 2.47 | Improved |
| add GMA | RAFT · Sintel test Clean | AEPE: 1.61 to 1.39 | Improved |
| add GMA content attention | RAFT · KITTI-15 test | Fl-all (%): 5.10 -> 5.15 | No clear effect |
| use position-only attention instead of content attention (ablation) | GMA · KITTI-15 test | Fl-all (%): 5.15 -> 4.93 | Improved |
| add positional attention to self-similarity attention (ablation) | GMA · Sintel train Clean | AEPE: 1.30 -> 1.33 | Hurt |
| learn scalar alpha instead of fixing it at 1 (ablation) | GMA · Sintel train Final | AEPE: 2.82 -> 2.74 | Improved |
| learn scalar alpha instead of fixing it at 1 (ablation) | GMA · Things Clean test | AEPE: 3.10 -> 3.14 | Hurt |
| concatenate aggregated features with local motion features instead of replacing (ablation) | GMA · Sintel train Final | AEPE: 2.79 -> 2.74 | Improved |
| concatenate aggregated features instead of replacing (ablation) | GMA · Chairs val | AEPE: 0.88 -> 0.79 | Improved |
Closest papers by meaning
- MirrorFlow: Exploiting Symmetries in Joint Optical Flow and Occlusion Estimation (ICCV 2017 · similarity 0.91)
- Unsupervised Learning of Multi-Frame Optical Flow with Occlusions (ECCV 2018 · similarity 0.90)
- Occlusions, Motion and Depth Boundaries with a Generic Network for Disparity, Optical Flow or Scene Flow Estimation (ECCV 2018 · similarity 0.90)
- OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation (CVPR 2024 · similarity 0.89)
- Dense Optical Tracking: Connecting the Dots (CVPR 2024 · similarity 0.87)
- OAMaskFlow: Occlusion-Aware Motion Mask for Scene Flow (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.
