Research map / Detection and segmentation / Weakly supervised detection and segmentation
Keep CALM and Improve Visual Feature Attribution
Jae Myung Kim, Junsuk Choe, Zeynep Akata, Seong Joon Oh
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 |
|---|---|---|---|
| replace CAM with CALM_EM attribution | ResNet50 · CUB class pairs attribution (mean) | mPxAP (%): 29.6 -> 34.3 | Improved |
| train with EM instead of marginal likelihood (ablation) | CALM · CUB class pairs attribution (mean) | mPxAP (%): 27.3 -> 34.3 | Improved |
| replace CAM with CALM_ML attribution | ResNet50 · CUB class pairs attribution (mean) | mPxAP (%): 29.6 -> 27.3 | Hurt |
| replace CAM-style classifier with CALM_EM | ResNet50 · ImageNet classification | accuracy (%): 74.5 -> 70.4 | Hurt |
| replace CAM-style classifier with CALM_ML | ResNet50 · CUB classification | accuracy (%): 70.6 -> 59.6 | Hurt |
| replace CAM-style classifier with CALM_EM | ResNet50 · CUB classification | accuracy (%): 70.6 -> 71.8 | No clear effect |
| add class-hierarchy superset aggregation (+Y) (ablation) | CALM_EM · CUB WSOL | MaxBoxAccV2 (%): 52.5 -> 65.4 | Improved |
| add class-hierarchy superset aggregation (+Y) (ablation) | CALM_ML · CUB WSOL | MaxBoxAccV2 (%): 61.3 -> 68.0 | Improved |
| add class-hierarchy superset aggregation (+Y) (ablation) | CAM · CUB WSOL | MaxBoxAccV2 (%): 61.1 -> 63.4 | Improved |
| add class-hierarchy superset aggregation (+Y) (ablation) | CALM_EM · ImageNet WSOL | MaxBoxAccV2 (%): 62.5 -> 62.8 | No clear effect |
| add class-hierarchy superset aggregation (+Y) (ablation) | CAM · ImageNet WSOL | MaxBoxAccV2 (%): 62.4 -> 60.6 | Hurt |
| add class-hierarchy superset aggregation (+Y) (ablation) | CALM_EM · OpenImages WSOL | PxAP (%): 62.7 -> 62.7 | No clear effect |
Closest papers by meaning
- Finer-CAM: Spotting the Difference Reveals Finer Details for Visual Explanation (CVPR 2025 · similarity 0.91)
- Fixing Localization Errors to Improve Image Classification (ECCV 2020 · similarity 0.89)
- DD-CAM: Minimal Sufficient Explanations for Vision Models Using Delta Debugging (NeurIPS 2026 · similarity 0.88)
- Benchmarking the Attribution Quality of Vision Models (NeurIPS 2024 · similarity 0.87)
- MoRe: Class Patch Attention Needs Regularization for Weakly Supervised Semantic Segmentation (AAAI 2025 · similarity 0.85)
- CAPE: CAM as a Probabilistic Ensemble for Enhanced DNN Interpretation (CVPR 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.
