Research map / Video / Video segmentation and tracking
Associating Objects and Their Effects in Video through Coordination Games
Erika Lu, Forrester Cole, Weidi Xie, Tali Dekel, William T. Freeman, Andrew Zisserman, Michael Rubinstein
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 |
|---|---|---|---|
| remove frame prediction task (K=0, loss on all frames) (ablation) | Ours pretrain · Shadows 2-object | IoU: 0.453 -> 0.638 | Improved |
| remove frame prediction task (K=0) (ablation) | Ours pretrain · Reflections 2-object | IoU: 0.440 -> 0.625 | Improved |
| mask 1 of 5 frames instead of 3 (ablation) | Ours pretrain · Shadows 2-object | IoU: 0.638 -> 0.679 | No clear effect |
| mask 4 of 5 frames instead of 3 (ablation) | Ours pretrain · Reflections 2-object | IoU: 0.625 -> 0.617 | No clear effect |
| inpainting vs random masking (ablation) | Ours pretrain · Shadows 2-object | IoU: 0.599 -> 0.615 | No clear effect |
| prediction vs random masking (ablation) | Ours pretrain · Reflections 2-object | IoU: 0.611 -> 0.596 | No clear effect |
| add test-time training with 200 iterations (estimated background) | Ours-A · Shadows 2-object | IoU: 0.607 -> 0.802 | Improved |
| use ground truth background in test-time training | Ours-B · Shadows 2-object | IoU: 0.802 -> 0.843 | Improved |
| pretrained feed-forward model without TTT vs single-video optimization | Ours-A · Reflections 4-object | IoU: 0.353 -> 0.569 | Improved |
| generalize from 2-object training to 4-object test | Ours-B · Shadows | IoU: 0.802 -> 0.752 | Mixed |
| fine-tune with larger shadow occlusion on real Dogwalk | Ours · Dogwalk real video | qualitative separation | Hurt |
Closest papers by meaning
- Generative Omnimatte: Learning to Decompose Video into Layers (CVPR 2025 · similarity 0.87)
- Free Supervision From Video Games (CVPR 2018 · similarity 0.87)
- Unsupervised Layered Image Decomposition Into Object Prototypes (ICCV 2021 · similarity 0.87)
- Omnimatte3D: Associating Objects and Their Effects in Unconstrained Monocular Video (CVPR 2023 · similarity 0.86)
- ObjectMover: Generative Object Movement with Video Prior (CVPR 2025 · similarity 0.85)
- Generative Image Layer Decomposition with Visual Effects (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.
