Research map / Label-efficient and robust learning / Domain adaptation and generalization
Leveraging Motion Priors in Videos for Improving Human Segmentation
Yu-Ting Chen, Wen-Yen Chang, Hai-Lun Lu, Tingfan Wu, Min Sun
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 |
|---|---|---|---|
| select motion priors with the learned policy instead of random (ablation) | U-Net segmenter · Gym-RGB->Gym-IR | IoU: 52.5% -> 55.6% | Improved |
| fine-tune on policy-selected motion priors from target video | U-Net segmenter · Gym-RGB->Gym-IR | IoU: 48.6% -> 55.6% | Improved |
| fine-tune on policy-selected motion priors | U-Net segmenter · Gym-RGB->UrbanStreet | IoU: 48.5% -> 51.2% | Improved |
| fine-tune on policy-selected motion priors | U-Net segmenter · Store-RGB->UrbanStreet | IoU: 61.7% -> 64.8% | Improved |
| learned policy instead of random selection (ablation) | U-Net segmenter · Gym-RGB->Multi-Scene-IR | IoU: 26.5% -> 30.5% | Improved |
| learned policy instead of random selection (ablation) | SegNet segmenter · Gym-RGB->Multi-Scene-IR | IoU: 37.2% -> 38.4% | Improved |
| learned policy instead of random selection (ablation) | SegNet segmenter · Store-RGB->Gym-IR | IoU: 24.8% -> 26.9% | Improved |
| learned policy instead of random selection (ablation) | U-Net segmenter · Store-RGB->Multi-Scene-IR | IoU: 30.2% -> 34.3% | Improved |
| learned policy instead of human selection of priors | U-Net segmenter · Store-RGB->Gym-IR | IoU: 32.5% -> 32.3% | No clear effect |
| add PAL on top of DSN | DSN · Gym-RGB->Multi-Scene-IR | IoU: 25.9% -> 35.8% | Improved |
| add PAL on top of DSN | DSN · Store-RGB->Gym-IR | IoU: 31.8% -> 36.4% | Improved |
| add PAL on top of DSN | DSN · Gym-RGB->Gym-IR | IoU: 54.3% -> 55.8% | Improved |
Closest papers by meaning
- Budget-Aware Deep Semantic Video Segmentation (CVPR 2017 · similarity 0.87)
- Weakly-Supervised Semantic Segmentation Using Motion Cues (ECCV 2016 · similarity 0.87)
- Reinforced active learning for image segmentation (ICLR 2020 · similarity 0.87)
- Learning to Segment Moving Objects (IJCV 2018 · similarity 0.87)
- Learning Segmentation from Point Trajectories (NeurIPS 2024 · similarity 0.86)
- Segment Any Motion in Videos (CVPR 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.
