Research map / Detection and segmentation / Weakly supervised detection and segmentation
Anti-Adversarially Manipulated Attributions for Weakly Supervised Semantic Segmentation and Object Localization
Jungbeom Lee, Eunji Kim, Jisoo Mok, Sungroh Yoon
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 |
|---|---|---|---|
| add adversarial climbing (ablation) | CutMix · CUB-200-2011 | Top-1 Loc: 57.6 -> 62.5 | Improved |
| add regularization on top of adversarial climbing (ablation) | CutMix + adversarial climbing · CUB-200-2011 | Top-1 Loc: 62.5 -> 64.9 | Improved |
| add regularization (lambda=7 vs 0) (ablation) | AdvCAM · PASCAL VOC 2012 train | seed mIoU: 50.43 -> 55.55 | Improved |
| aggregate CAMs over all iterations instead of last iteration (ablation) | AdvCAM · PASCAL VOC 2012 train | seed mIoU: 54.0 -> 55.6 | Improved |
| add adversarial climbing (ablation) | Grad-CAM++ · PASCAL VOC 2012 train | seed mIoU: 47.7 -> 50.7 | Improved |
| add adversarial climbing (ablation) | Chang et al. · PASCAL VOC 2012 train | seed mIoU: 50.9 -> 53.7 | Improved |
| add adversarial climbing (ablation) | SEAM · PASCAL VOC 2012 train | seed mIoU: 55.4 -> 58.6 | Improved |
| add adversarial climbing (ablation) | IRN · PASCAL VOC 2012 train | seed mIoU: 48.8 -> 55.6 | Improved |
| add salient object detector (ablation) | AdvCAM · PASCAL VOC 2012 train | seed mIoU: 55.6 -> 60.8 | Improved |
| add CRF post-processing | AdvCAM · PASCAL VOC 2012 train | seed mIoU: 55.6 -> 62.1 | Improved |
| vary masking threshold tau from 0.3 to 0.7 (ablation) | AdvCAM · PASCAL VOC 2012 train | seed mIoU | No clear effect |
| apply adversarial climbing to small-object classes (ablation) | AdvCAM vs CAM · MS COCO 2014 smallest 10 classes | recall: 73.0 -> 64.9 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CutMix — CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features (ICCV 2019) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Anti-Adversarially Manipulated Attributions for Weakly and Semi-Supervised Semantic Segmentation (CVPR 2021 · similarity 0.97)
- Unlocking the Potential of Ordinary Classifier: Class-Specific Adversarial Erasing Framework for Weakly Supervised Semantic Segmentation (ICCV 2021 · similarity 0.88)
- Adversarial Complementary Learning for Weakly Supervised Object Localization (CVPR 2018 · similarity 0.87)
- Weakly Supervised Semantic Segmentation via Adversarial Learning of Classifier and Reconstructor (CVPR 2023 · similarity 0.87)
- BBAM: Bounding Box Attribution Map for Weakly Supervised Semantic and Instance Segmentation (CVPR 2021 · similarity 0.86)
- Object Region Mining With Adversarial Erasing: A Simple Classification to Semantic Segmentation Approach (CVPR 2017 · 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.
