Research map / Label-efficient and robust learning / Domain adaptation and generalization
Toward Real-World High-Precision Image Matting and Segmentation
Haipeng Zhou, Zhaohu Xing, Hongqiu Wang, Jun Ma, Ping Li, Lei Zhu
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 |
|---|---|---|---|
| train on synthetic instead of real backgrounds | MODNet · P3M-500-P | SAD: 10.33 -> 23.41 | Hurt |
| add vanilla feature distillation (ablation) | FCLM · RefMatte-RW | SAD: 26.34 -> 23.48 | Improved |
| add depth-aware foreground distillation from both images (ablation) | FCLM · RefMatte-RW | SAD: 23.48 -> 22.18 | Improved |
| add background distillation (ablation) | FCLM · RefMatte-RW | SAD: 22.18 -> 21.31 | Improved |
| add domain alignment (adversarial + OT) (ablation) | FCLM · RefMatte-RW | SAD: 24.33 -> 21.31 | Improved |
| add L_OT to L_adv (ablation) | FCLM · RefMatte-RW | Grad: 23.01 -> 19.38 | Improved |
| use box prompts instead of text prompts (ablation) | FCLM · RefMatte-RW | SAD: 21.31 -> 16.37 | Improved |
| use point prompts instead of text prompts (ablation) | FCLM · RefMatte-RW | SAD: 21.31 -> 18.46 | Improved |
| use FCLM instead of MAM on connectivity | FCLM · RefMatte-RW | Conn: 2.83 -> 3.33 | Hurt |
Closest papers by meaning
- Privileged Prior Information Distillation for Image Matting (AAAI 2024 · similarity 0.88)
- Training Matting Models Without Alpha Labels (AAAI 2025 · similarity 0.87)
- FOCUS: Towards Universal Foreground Segmentation (AAAI 2025 · similarity 0.86)
- Unifying Automatic and Interactive Matting with Pretrained ViTs (CVPR 2024 · similarity 0.86)
- Early Semantic Grounding in Image Editing Models for Zero-Shot Referring Image Segmentation (NeurIPS 2026 · similarity 0.86)
- In-Context Matting (CVPR 2024 · 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.
