Research map / Detection and segmentation / Open-set and few-shot detection
ZIM: Zero-Shot Image Matting for Anything
Beomyoung Kim, Chanyong Shin, Joonhyun Jeong, Hyungsik Jung, Se-Yun Lee, Sewhan Chun, Dong-Hyun Hwang, Joonsang Yu
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 |
|---|---|---|---|
| add Spatial Generalization Augmentation to label converter (ablation) | label converter · MicroMat-3K Fine-grained | SAD: 3.324 -> 2.440 | Improved |
| add Selective Transformation Learning to label converter (ablation) | label converter · MicroMat-3K Fine-grained | SAD: 3.324 -> 3.153 | Improved |
| add both SGA and STL to label converter (ablation) | label converter · MicroMat-3K Coarse-grained | SAD: 0.716 -> 0.281 | Improved |
| add prompt-aware masked attention to ZIM (ablation) | ZIM · MicroMat-3K Fine-grained (box) | MSE: 2.718 -> 2.504 | Improved |
| add hierarchical pixel decoder to ZIM (ablation) | ZIM · MicroMat-3K Fine-grained (box) | Grad: 6.516 -> 5.401 | Improved |
| add both attention and decoder to ZIM (ablation) | ZIM · MicroMat-3K Fine-grained (box) | MSE: 2.718 -> 1.893 | Improved |
| apply masked attention in T2I only versus no masked attention (ablation) | ZIM · MicroMat-3K Fine-grained (box) | MSE: 2.094 -> 1.893 | Improved |
| apply masked attention in I2T layer only versus none (ablation) | ZIM · MicroMat-3K Fine-grained (box) | MSE: 2.094 -> 2.658 | Hurt |
| also apply masked attention in I2T in addition to T2I (ablation) | ZIM · MicroMat-3K Fine-grained (box) | MSE: 1.893 -> 1.997 | Hurt |
| train ZIM on SA1B-Matte instead of public matting datasets (ablation) | ZIM · MicroMat-3K Fine-grained (box) | MSE: 38.332 -> 1.893 | Improved |
| train Matting-Any on SA1B-Matte instead of public matting data (ablation) | Matting-Any · MicroMat-3K Fine-grained (box) | MSE: 68.372 -> 12.267 | Improved |
| replace SAM with ZIM in Inpaint Anything masks | Inpainting Anything · COCO inpainting | CLIP Distance: 68.17 -> 73.11 | Improved |
Closest papers by meaning
- Segment and Matte Anything in a Unified Model (AAAI 2026 · similarity 0.90)
- Matting Anything 2: Towards Video Matting for Anything (ICLR 2026 · similarity 0.87)
- AlignZeg: Mitigating Objective Misalignment for Zero-shot Semantic Segmentation (ECCV 2024 · similarity 0.87)
- Unifying Automatic and Interactive Matting with Pretrained ViTs (CVPR 2024 · similarity 0.86)
- In-Context Matting (CVPR 2024 · similarity 0.85)
- DiffCut: Catalyzing Zero-Shot Semantic Segmentation with Diffusion Features and Recursive Normalized Cut (NeurIPS 2024 · similarity 0.85)
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.
