Research map / Image generation / Autoregressive and latent generation
High-Resolution Deep Image Matting
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 |
|---|---|---|---|
| Add CPC cross-patch context modeling (ablation) | HDMatt encoder-decoder · AIM test set | SAD: 41.1 -> 33.5 | Helped |
| Replace standard non-local operation with trimap-guided TGNL (ablation) | HDMatt with CPC · AIM test set | SAD: 35.6 -> 33.5 | Helped |
| Use patch size 480 instead of 320 (ablation) | HDMatt with CPC and TGNL · AIM test set | SAD: 33.5 -> 33.0 | No clear effect |
| Use patch size 640 instead of 320 (ablation) | HDMatt with CPC and TGNL · AIM test set | SAD: 33.5 -> 33.1 | No clear effect |
| Train with mixed patch sizes 320, 480, and 640 instead of 320 (ablation) | HDMatt with CPC and TGNL · AIM test set | SAD: 33.5 -> 32.8 | Helped |
| Use one context patch instead of three (ablation) | HDMatt with CPC and TGNL · AIM test set | SAD: 33.5 -> 34.9 | Hurt |
| Use five context patches instead of three (ablation) | HDMatt with CPC and TGNL · AIM test set | SAD: 33.5 -> 33.3 | No clear effect |
| Use seven context patches instead of three (ablation) | HDMatt with CPC and TGNL · AIM test set | SAD: 33.5 -> 33.2 | No clear effect |
| Use all context patches instead of three (ablation) | HDMatt with CPC and TGNL · AIM test set | SAD: 33.5 -> 32.2 | Helped |
Closest papers by meaning
- Deep Image Matting (CVPR 2017 · similarity 0.91)
- Revisiting Context Aggregation for Image Matting (ICML 2024 · similarity 0.89)
- Training Matting Models Without Alpha Labels (AAAI 2025 · similarity 0.88)
- In-Context Matting (CVPR 2024 · similarity 0.87)
- Privileged Prior Information Distillation for Image Matting (AAAI 2024 · similarity 0.86)
- Memory Efficient Matting with Adaptive Token Routing (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.