Research map / Detection and segmentation / Weakly supervised detection and segmentation
Extracting Class Activation Maps From Non-Discriminative Features As Well
Read the paperCheck 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 |
|---|---|---|---|
| Replace CAM with LPCAM in IRN | IRN · PASCAL VOC 2012 val, DeepLabV2 | mIoU: 63.5 -> 68.6 | Helped |
| Replace CAM with LPCAM in AMN | AMN · PASCAL VOC 2012 val, DeepLabV2 | mIoU: 69.5 -> 70.1 | Helped |
| Replace CAM with LPCAM in IRN | IRN · MS COCO 2014 val, DeepLabV2 | mIoU: 42.0 -> 44.5 | Helped |
| Replace CAM with LPCAM in AMN | AMN · MS COCO 2014 val, DeepLabV2 | mIoU: 44.7 -> 45.5 | Helped |
| Add foreground local prototypes to CAM (ablation) | CAM seed masks · PASCAL VOC 2012 train seed masks | recall: 65.2 -> 76.6 | Helped |
| Add context prototypes and subtract context similarity maps (ablation) | LPCAM-F seed masks · PASCAL VOC 2012 train seed masks | mIoU: 52.1 -> 54.9 | Helped |
| Add context prototypes and subtract context similarity maps (ablation) | LPCAM-F seed masks · MS COCO 2014 train seed masks | mIoU: 33.9 -> 35.4 | Helped |
| Use foreground local prototypes instead of conventional classifier CAM (ablation) | CAM seed masks · PASCAL VOC 2012 train seed masks | false positives: 26.5 -> 33.1 | Hurt |
| Use foreground local prototypes instead of conventional classifier CAM (ablation) | CAM seed masks · MS COCO 2014 train seed masks | false positives: 45.7 -> 49.7 | Hurt |
| Use hierarchical clustering instead of K-Means (ablation) | LPCAM · PASCAL VOC 2012 train seed masks | mIoU: 54.9 -> 55.1 | Helped |
| Use Euclidean similarity instead of cosine similarity (ablation) | LPCAM · PASCAL VOC 2012 train seed masks | mIoU: 54.9 -> 54.4 | Hurt |
| Increase K-Means cluster count from 12 to 20 (ablation) | LPCAM · PASCAL VOC 2012 train seed masks | mIoU: drops by only 0.7 percentage points | Hurt |
Closest papers by meaning
- Finer-CAM: Spotting the Difference Reveals Finer Details for Visual Explanation (CVPR 2025 · similarity 0.90)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.87)
- MoRe: Class Patch Attention Needs Regularization for Weakly Supervised Semantic Segmentation (AAAI 2025 · similarity 0.87)
- PSDPM: Prototype-based Secondary Discriminative Pixels Mining for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.87)
- Class Tokens Infusion for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.86)
- CAPE: CAM as a Probabilistic Ensemble for Enhanced DNN Interpretation (CVPR 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.