Research map / Detection and segmentation / Weakly supervised detection and segmentation
Class Re-Activation Maps for Weakly-Supervised Semantic Segmentation
Zhaozheng Chen, Tan Wang, Xiongwei Wu, Xian-Sheng Hua, Hanwang Zhang, Qianru Sun
Check 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 |
|---|---|---|---|
| Reactivate CAM using the BCE-initialized model and an additional SCE-trained classifier | CAM pseudo-mask generation · PASCAL VOC 2012 train pseudo masks | mIoU (%): 48.8 -> 54.8 | Improved |
| Reactivate CAM using the BCE-initialized model and an additional SCE-trained classifier | CAM pseudo-mask generation · MS COCO 2014 train pseudo masks | mIoU (%): 33.1$^*$ -> 34.6 | Improved |
| Use ReCAM seeds instead of CAM seeds before IRN refinement | IRN-refined pseudo masks · PASCAL VOC 2012 train pseudo masks | mIoU (%): 66.3 -> 70.9 | Improved |
| Use ReCAM seeds instead of CAM seeds before IRN refinement | IRN-refined pseudo masks · MS COCO 2014 train pseudo masks | mIoU (%): 42.4 -> 44.1 | Improved |
| Use ReCAM seeds instead of CAM seeds | DeepLabV2 segmentation · PASCAL VOC 2012 val | mIoU (%): 63.5 -> 68.7 | Improved |
| Use ReCAM seeds instead of CAM seeds | UperNet-Swin segmentation · PASCAL VOC 2012 val | mIoU (%): 65.9 -> 70.9 | Improved |
| Use SCE alone instead of BCE for multi-label FC1 training (ablation) | CAM classifier without FC2 · PASCAL VOC 2012 pseudo masks | mIoU (%): 48.8 -> 44.6 | Hurt |
| Use SCE for single-label images and BCE for multi-label images (ablation) | CAM classifier without FC2 · PASCAL VOC 2012 pseudo masks | mIoU (%): 48.8 -> 49.4 | Improved |
| Use SCE for single-label images and BCE for multi-label images (ablation) | CAM classifier without FC2 · MS COCO 2014 pseudo masks | mIoU (%): 33.1 -> 33.4 | Improved |
| Extract ReCAM using FC1 weights rather than multiplied FC1 and FC2 weights (ablation) | ReCAM · PASCAL VOC 2012 pseudo masks | mIoU (%): 54.8 (rp.) -> 52.1 | Hurt |
| Extract ReCAM using FC2 weights rather than multiplied FC1 and FC2 weights (ablation) | ReCAM · PASCAL VOC 2012 pseudo masks | mIoU (%): 54.8 (rp.) -> 54.1 | Hurt |
| Extract ReCAM using FC1 weights rather than multiplied FC1 and FC2 weights (ablation) | ReCAM · MS COCO 2014 pseudo masks | mIoU (%): 34.0 -> 34.6 (rp.) | Improved |
Closest papers by meaning
- Class Tokens Infusion for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.91)
- Know Your Attention Maps: Class-specific Token Masking for Weakly Supervised Semantic Segmentation (ICCV 2025 · similarity 0.91)
- Pseudo-Mask Matters in Weakly-Supervised Semantic Segmentation (ICCV 2021 · similarity 0.91)
- Activation Modulation and Recalibration Scheme for Weakly Supervised Semantic Segmentation (AAAI 2022 · similarity 0.90)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.90)
- Weakly Supervised Semantic Segmentation via Progressive Confidence Region Expansion (CVPR 2025 · 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.
