Research map / Detection and segmentation / Open-set and few-shot detection
Unsupervised Multi-Object Segmentation by Predicting Probable Motion Patterns
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 |
|---|---|---|---|
| use affine instead of translation-only motion model (ablation) | Ours · MOVi-A | FG-ARI: 66.03 -> 83.48 | Helped |
| use affine instead of translation-only motion model (ablation) | Ours · MOVi-C | mIoU: 32.23 -> 35.67 | Helped |
| use ground-truth optical flow instead of SMURF-estimated flow (ablation) | Ours · MOVi-A | mIoU: 26.3 -> 72.61 | Helped |
| use ground-truth optical flow instead of SMURF-estimated flow (ablation) | Ours · MOVi-C | FG-ARI: 61.21 -> 58.59 | No clear effect |
| add motion-based loss to Slot Attention (ablation) | Slot Attention · ClevrTex | mIoU: 21.44 -> 38.12 | Helped |
| add motion-based loss to GNM (ablation) | GNM · ClevrTex | FG-ARI: 18.01 -> 63.84 | Helped |
| add post-processing (connected components) to Ours (ablation) | Ours · ClevrTex | mIoU: 55.07 -> 77.67 | Helped |
| add post-processing to baseline methods | baselines · CLEVR/ClevrTex | FG-ARI/mIoU | Mixed |
| add warp loss and Swin backbone (ablation) | Ours · MOVi-C | mIoU: 37.26 -> 52.17 | Helped |
| add warp loss to ResNet-18 model (ablation) | Ours · KITTI | FG-ARI: 50.8 -> 51.9 | Helped |
| use Swin backbone with warp loss on KITTI (ablation) | Ours · KITTI | FG-ARI: 51.9 -> 58.3 | Helped |
| train image-only Slot Attention on MovingClevrTex videos without the motion loss (ablation) | Slot Attention · ClevrTex | FG-ARI: 62.40 -> 61.84 | No clear effect |
Closest papers by meaning
- Learning Features by Watching Objects Move (CVPR 2017 · similarity 0.91)
- Learning Motion Patterns in Videos (CVPR 2017 · similarity 0.91)
- Learning Segmentation from Point Trajectories (NeurIPS 2024 · similarity 0.90)
- unMORE: Unsupervised Multi-Object Segmentation via Center-Boundary Reasoning (ICML 2025 · similarity 0.89)
- Segment Any Motion in Videos (CVPR 2025 · similarity 0.87)
- M^3-VOS: Multi-Phase, Multi-Transition, and Multi-Scenery Video Object Segmentation (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.