Research map / Detection and segmentation / Weakly supervised detection and segmentation
Knowledge Transfer with Simulated Inter-Image Erasing for Weakly Supervised Semantic Segmentation
Tao Chen, Xiruo Jiang, Gensheng Pei, Zeren Sun, Yucheng Wang, Yazhou Yao
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 simulated inter-image erasing (SIE) to the baseline (ablation) | KTSE baseline (GPP) · PASCAL VOC 2012 train (seed) | mIoU: 54.2 -> 57.1 | Improved |
| add self-supervised regularization (SSR) to the baseline (ablation) | KTSE baseline (GPP) · PASCAL VOC 2012 train (seed) | mIoU: 54.2 -> 57.4 | Improved |
| add multi-granularity alignment (MGA) to the baseline (ablation) | KTSE baseline (GPP) · PASCAL VOC 2012 train (seed) | mIoU: 54.2 -> 60.2 | Improved |
| combine SIE with SSR (ablation) | KTSE · PASCAL VOC 2012 train (seed) | mIoU: 57.1 -> 63.8 | Improved |
| add MGA on top of SIE and SSR (ablation) | KTSE · PASCAL VOC 2012 train (seed) | mIoU: 63.8 -> 67.0 | Improved |
| use global alignment on CAM features instead of rigid classification guidance (ablation) | MGA · PASCAL VOC 2012 train (seed) | mIoU: 55.6 -> 58.7 | Improved |
| align GPP embedding features instead of CAM features (ablation) | MGA · PASCAL VOC 2012 train (seed) | mIoU: 58.7 -> 54.9 | Hurt |
| add pixel-level local alignment to global alignment (ablation) | MGA · PASCAL VOC 2012 train (seed) | mIoU: 58.7 -> 60.2 | Improved |
| KTSE vs CutMix-based augmentation method CDA | KTSE · PASCAL VOC 2012 val (ResNet) | mIoU: 66.1 -> 73.0 | Improved |
| KTSE vs CutMix-based augmentation method CDA | KTSE · COCO val (ResNet) | mIoU: 33.2 -> 45.9 | Improved |
| SIE alone causes more under-activation for larger objects (bidirectional alignment), mitigated by SSR | SIE · PASCAL VOC 2012 train (qualitative) | localization maps | Mixed |
| add IRN refinement to KTSE seeds | KTSE · PASCAL VOC 2012 train | mIoU: 67.0 -> 73.8 | Improved |
Closest papers by meaning
- Adversarial Erasing Framework via Triplet with Gated Pyramid Pooling Layer for Weakly Supervised Semantic Segmentation (ECCV 2022 · similarity 0.90)
- Object Region Mining With Adversarial Erasing: A Simple Classification to Semantic Segmentation Approach (CVPR 2017 · similarity 0.89)
- Erasing Integrated Learning: A Simple Yet Effective Approach for Weakly Supervised Object Localization (CVPR 2020 · similarity 0.89)
- Unlocking the Potential of Ordinary Classifier: Class-Specific Adversarial Erasing Framework for Weakly Supervised Semantic Segmentation (ICCV 2021 · similarity 0.89)
- ECS-Net: Improving Weakly Supervised Semantic Segmentation by Using Connections Between Class Activation Maps (ICCV 2021 · similarity 0.89)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · 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.
