Research map / Detection and segmentation / Weakly supervised detection and segmentation
Multi-Class Token Transformer for Weakly Supervised Semantic 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 multiple class-specific tokens instead of TS-CAM's class-agnostic attention (ablation) | PASCAL VOC train localization · PASCAL VOC train | mIoU: 41.3 -> 47.2 | Helped |
| Add PatchAffinity refinement to MCTformer-V1 attention (ablation) | MCTformer-V1 · PASCAL VOC train | mIoU: 47.2 -> 55.2 | Helped |
| Add PatchCAM and fuse it with class-specific transformer attention (ablation) | MCTformer-V2 · PASCAL VOC train | mIoU: 55.2 -> 58.2 | Helped |
| Add PatchAffinity refinement to fused attention and PatchCAM (ablation) | MCTformer-V2 · PASCAL VOC train | mIoU: 58.2 -> 61.7 | Helped |
| Add PatchAffinity refinement to MCTformer-V1 attention (ablation) | MCTformer-V1 · PASCAL VOC val | mIoU: 55.6 -> 58.8 | Helped |
| Add PatchCAM and fuse it with class-specific transformer attention (ablation) | MCTformer-V2 · PASCAL VOC val | mIoU: 58.8 -> 61.1 | Helped |
| Add PatchAffinity refinement to fused attention and PatchCAM (ablation) | MCTformer-V2 · PASCAL VOC val | mIoU: 61.1 -> 62.6 | Helped |
| Replace average pooling with a fully-connected layer for class-token scores (ablation) | MCTformer-V1 · PASCAL VOC train | mIoU: 47.2 -> 41.5 | Hurt |
| Replace average pooling with max pooling for class-token scores (ablation) | MCTformer-V1 · PASCAL VOC train | mIoU: 47.2 -> 26.8 | Hurt |
| Fuse attentions from the last three transformer layers (ablation) | MCTformer-V1 · PASCAL VOC train | mIoU: 47.2 | Helped |
| Use MCTformer-generated pseudo labels rather than the compared prior pseudo-label method | PASCAL VOC train seed and mask · PASCAL VOC train | mIoU: 57.4 -> 61.7; 67.8 -> 69.1 | Helped |
| Use MCTformer pseudo labels for semantic segmentation instead of the strongest image-level-label baseline | PASCAL VOC 2012 · PASCAL VOC 2012 val | mIoU: 68.4 -> 71.9 | Helped |
Closest papers by meaning
- Learning Multi-Modal Class-Specific Tokens for Weakly Supervised Dense Object Localization (CVPR 2023 · similarity 0.93)
- MoRe: Class Patch Attention Needs Regularization for Weakly Supervised Semantic Segmentation (AAAI 2025 · similarity 0.91)
- Class Tokens Infusion for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.91)
- Token Contrast for Weakly-Supervised Semantic Segmentation (CVPR 2023 · similarity 0.90)
- Discriminative Region Suppression for Weakly-Supervised Semantic Segmentation (AAAI 2021 · similarity 0.88)
- Weakly Supervised Semantic Segmentation via Progressive Confidence Region Expansion (CVPR 2025 · similarity 0.87)
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.