Research map / Video / Video-language understanding
SAM2Matting: Generalized Image and Video Matting
Ruiqi Shen, Guangquan Jie, Chang Liu, Henghui Ding
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 |
|---|---|---|---|
| use learned ROI Detector instead of morphological ROI (ablation) | SAM2Matting (SAM2.1-B+) · V-HIM60-Hard | MAD: 29.82 -> 18.20 | Improved |
| use learned ROI Detector instead of raw mask as ROI (ablation) | SAM2Matting (SAM2.1-B+) · V-HIM60-Hard | MAD: 20.07 -> 18.20 | Improved |
| use learned ROI Detector instead of raw mask as ROI (ablation) | SAM2Matting (SAM2.1-B+) · V-HIM60-Hard | dtSSD: 5.50 -> 5.10 | Improved |
| add progressive multi-scale cascade in alpha predictor (ablation) | SAM2Matting (SAM2.1-B+) · V-HIM60-Hard | MAD: 19.43 -> 18.20 | Improved |
| add matte-mask consistency loss (ablation) | SAM2Matting (SAM2.1-B+) · V-HIM60-Hard | MAD: 18.65 -> 18.20 | Improved |
| add smoothness loss (ablation) | SAM2Matting (SAM2.1-B+) · V-HIM60-Hard | MAD: 18.26 -> 18.20 | No clear effect |
| fine-tune on V-HIM2K5 video matting data (ablation) | SAM2Matting (SAM2.1-B+) · V-HIM60-Hard | MAD: 18.20 -> 17.90 | Improved |
| fine-tune on V-HIM2K5 video matting data (ablation) | SAM2Matting (SAM2.1-B+) · AM-2K test | MAD: 4.90 -> 5.23 | Hurt |
| fine-tune on V-HIM2K5 video matting data (ablation) | SAM2Matting (SAM2.1-B+) · VideoMatte-SD | MAD: 4.83 -> 4.85 | No clear effect |
| use SAM3 tracker instead of SAM2.1-B+ (ablation) | SAM2Matting · V-HIM60-Hard | MAD: 18.20 -> 14.37 | Improved |
| use SAM2Matting with same SAM2.1-B+ tracker as baselines instead of Matte Anything (ablation) | SAM2Matting (SAM2.1-B+) · P3M-500-NP | MAD: 6.00 -> 3.81 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- MatAnyone — MatAnyone: Stable Video Matting with Consistent Memory Propagation (CVPR 2025) · better on 3 of 3 shared results
- MatAnyone2 — MatAnyone 2: Scaling Video Matting via a Learned Quality Evaluator (CVPR 2026) · better on 2 of 2 shared results
- MaGGIe — MaGGIe: Masked Guided Gradual Human Instance Matting (CVPR 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- VideoMaMa: Mask-Guided Video Matting via Generative Prior (CVPR 2026 · similarity 0.92)
- $\alpha$Matte4K & $\mu$Matting: Dataset and Model for Ultra-Micro Precision Alpha Video Matting (CVPR 2026 · similarity 0.91)
- MatAnyone: Stable Video Matting with Consistent Memory Propagation (CVPR 2025 · similarity 0.90)
- Training Matting Models Without Alpha Labels (AAAI 2025 · similarity 0.89)
- MaGGIe: Masked Guided Gradual Human Instance Matting (CVPR 2024 · similarity 0.88)
- In-Context Matting (CVPR 2024 · similarity 0.88)
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.
