Research map / Detection and segmentation / Weakly supervised detection and segmentation
Object Region Mining With Adversarial Erasing: A Simple Classification to Semantic Segmentation Approach
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 |
|---|---|---|---|
| progressively add mined object regions from AE steps 1–3 (ablation) | AE segmentation supervision · PASCAL VOC 2012 val | mIoU: 44.9 -> 49.5 -> 50.9 | Helped |
| add object regions mined at AE step 4 (ablation) | AE segmentation supervision · PASCAL VOC 2012 val | mIoU: 50.9 -> 48.8 | Hurt |
| add online PSL auxiliary segmentation supervision (ablation) | AE segmentation training · PASCAL VOC 2012 val | mIoU: 50.9 -> 54.1 | Helped |
| add one more training round using predicted masks after CRF (ablation) | AE with PSL · PASCAL VOC 2012 val | mIoU: 54.1 -> 55.0 | Helped |
| use ground-truth image-level labels instead of predicted classification confidences for training-time prohibition (ablation) | PSL · PASCAL VOC 2012 val | mIoU: -0.6% | Hurt |
| remove image-level classification confidences from PSL (ablation) | PSL · PASCAL VOC 2012 val | mIoU: -1% | Hurt |
| use ground-truth image-level labels to weight test-time segmentation score maps (ablation) | PSL++ · PASCAL VOC 2012 val | mIoU: 55.0 -> 56.1 | Helped |
| perform additional training rounds after PSL++ (ablation) | AE with PSL · PASCAL VOC 2012 val | mIoU | No clear effect |
| use AE-PSL rather than the strongest image-level-label baseline AF-MCG | PASCAL VOC 2012 val · PASCAL VOC 2012 val | mIoU: 54.3 -> 55.0 | Helped |
| use AE-PSL rather than AF-MCG | PASCAL VOC 2012 test · PASCAL VOC 2012 test | mIoU: 55.5 -> 55.7 | Helped |
Closest papers by meaning
- Erasing Integrated Learning: A Simple Yet Effective Approach for Weakly Supervised Object Localization (CVPR 2020 · similarity 0.89)
- Weakly Supervised Semantic Segmentation via Adversarial Learning of Classifier and Reconstructor (CVPR 2023 · similarity 0.88)
- Weakly-Supervised Semantic Segmentation by Iteratively Mining Common Object Features (CVPR 2018 · similarity 0.87)
- Revisiting Dilated Convolution: A Simple Approach for Weakly- and Semi-Supervised Semantic Segmentation (CVPR 2018 · similarity 0.87)
- Non-Salient Region Object Mining for Weakly Supervised Semantic Segmentation (CVPR 2021 · similarity 0.87)
- Self-Erasing Network for Integral Object Attention (NeurIPS 2018 · 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.