Research map / Detection and segmentation / CNN features
Masked-Attention Mask Transformer for Universal Image Segmentation
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 masked attention instead of global cross-attention (ablation) | Mask2Former · COCO instance val2017 | AP: 37.8 -> 43.7 | Helped |
| Use masked attention instead of global cross-attention (ablation) | Mask2Former · COCO panoptic val2017 | PQ: 47.1 -> 51.9 | Helped |
| Use masked attention instead of global cross-attention (ablation) | Mask2Former · ADE20K val | mIoU: 45.5 -> 47.2 | Helped |
| Add high-resolution features through the efficient multi-scale strategy (ablation) | Mask2Former · COCO instance val2017 | AP: 41.5 -> 43.7 | Helped |
| Add high-resolution features through the efficient multi-scale strategy (ablation) | Mask2Former · COCO panoptic val2017 | PQ: 50.2 -> 51.9 | Helped |
| Add high-resolution features through the efficient multi-scale strategy (ablation) | Mask2Former · ADE20K val | mIoU: 46.1 -> 47.2 | Helped |
| Make query features learnable rather than zero-initialized (ablation) | Mask2Former · COCO instance val2017 | AP: 42.9 -> 43.7 | Helped |
| Make query features learnable rather than zero-initialized (ablation) | Mask2Former · COCO panoptic val2017 | PQ: 51.2 -> 51.9 | Helped |
| Make query features learnable rather than zero-initialized (ablation) | Mask2Former · ADE20K val | mIoU: 45.4 -> 47.2 | Helped |
| Put masked cross-attention before self-attention (ablation) | Mask2Former · COCO instance val2017 | AP: 43.2 -> 43.7 | Helped |
| Remove dropout from the decoder (ablation) | Mask2Former · COCO instance val2017 | AP: 43.0 -> 43.7 | Helped |
| Remove dropout from the decoder (ablation) | Mask2Former · ADE20K val | mIoU: 47.2 -> 47.2 | No clear effect |
Closest papers by meaning
- MP-Former: Mask-Piloted Transformer for Image Segmentation (CVPR 2023 · similarity 0.92)
- Per-Pixel Classification is Not All You Need for Semantic Segmentation (NeurIPS 2021 · similarity 0.92)
- PEM: Prototype-based Efficient MaskFormer for Image Segmentation (CVPR 2024 · similarity 0.90)
- Mask DINO: Towards a Unified Transformer-Based Framework for Object Detection and Segmentation (CVPR 2023 · similarity 0.89)
- Rethinking Decoders for Transformer-based Semantic Segmentation: A Compression Perspective (NeurIPS 2024 · similarity 0.88)
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 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.