Research map / Image generation / Autoregressive and latent generation
Boosting Robustness of Image Matting With Context Assembling and Strong Data Augmentation
Check 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 the proposed M7 framework instead of FBA+TTA | Image matting · Composition-1k test | SAD: 25.8 -> 22.87 | Helped |
| Use the proposed M7 framework instead of FBA+TTA | Image matting · Composition-1k test | Grad: 10.6 -> 7.74 | Helped |
| Use the proposed M7 model | Image matting · AIM-500 | SAD: 19.05 -> 13.97 | Helped |
| Replace an MLP decoder with a convolution decoder (ablation) | B1 on Composition-1k · Composition-1k test | SAD: 39.55 -> 29.85 | Helped |
| Add low-level skip fusion (ablation) | B2 on Composition-1k · Composition-1k test | SAD: 29.85 -> 28.94 | Helped |
| Add Laplacian loss and gradient-penalty loss (ablation) | B3 · Composition-1k test | SAD: 28.94 -> 25.86 | Helped |
| Add the low-level feature assembling attention block (LFA) (ablation) | N3 on Composition-1k · Composition-1k test | SAD: 25.86 -> 25.70 | Helped |
| Add foreground-only augmentation (AF) (ablation) | N3 · AIM-500 | SAD: 16.08 -> 16.38 | Hurt |
| Add separate foreground and background augmentation (AFB) (ablation) | N3 · AIM-500 | SAD: 16.08 -> 13.46 | Helped |
| Add AF, AFB, and composited-image augmentation (AC) (ablation) | N3 · AIM-500 | SAD: 16.08 -> 14.18 | Helped |
| Apply extra DA without changing alpha ground truths (ablation) | DIM-trained model · Composition-1k test | SAD: 32.65 -> 35.79 | Hurt |
| Modify alpha ground truths to match extra DA (ablation) | DIM-trained model using DA · Composition-1k test | SAD: 35.79 -> 34.00 | Helped |
Closest papers by meaning
- Deep Image Matting (CVPR 2017 · similarity 0.91)
- Training Matting Models Without Alpha Labels (AAAI 2025 · similarity 0.91)
- Improved Image Matting via Real-Time User Clicks and Uncertainty Estimation (CVPR 2021 · similarity 0.90)
- High-Resolution Deep Image Matting (AAAI 2021 · similarity 0.90)
- Revisiting Context Aggregation for Image Matting (ICML 2024 · similarity 0.90)
- In-Context Matting (CVPR 2024 · similarity 0.90)
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.