Research map / Detection and segmentation / Weakly supervised detection and segmentation
Background-Aware Pooling and Noise-Aware Loss for Weakly-Supervised Semantic Segmentation
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 |
|---|---|---|---|
| Use BAP instead of GAP to aggregate foreground features (ablation) | pseudo-label generation on PASCAL VOC 2012 · PASCAL VOC 2012 val | pseudo-label mIoU: 76.1 -> 79.2 | Helped |
| Add the adaptive background attention unary u_0 rather than use a background-class CAM (ablation) | BAP pseudo-label generation · PASCAL VOC 2012 val | pseudo-label mIoU: 77.8 -> 79.2 | Helped |
| Use retrieval-based pseudo-labels Y_ret instead of DenseCRF labels Y_crf (ablation) | BAP pseudo-label generation · PASCAL VOC 2012 val | pseudo-label mIoU: 79.2 -> 69.9 | Hurt |
| Add confidence-weighted cross-entropy for disagreeing pseudo-label regions (ablation) | DeepLab-V1 segmentation training · PASCAL VOC 2012 val | mIoU before/after DenseCRF: 61.8 / 67.5 -> 62.4 / 68.1 | Helped |
| Use entropy regularization on disagreeing pseudo-label regions (ablation) | DeepLab-V1 segmentation training · PASCAL VOC 2012 val | mIoU before/after DenseCRF: 61.8 / 67.5 -> 61.4 / 67.3 | Hurt |
| Use bootstrapping on disagreeing pseudo-label regions (ablation) | DeepLab-V1 segmentation training · PASCAL VOC 2012 val | mIoU before/after DenseCRF: 61.8 / 67.5 -> 61.9 / 67.6 | Helped |
| Train with both DenseCRF and retrieval pseudo-labels plus NAL (ablation) | DeepLab-V1 segmentation · PASCAL VOC 2012 val | mIoU: 67.8 -> 68.1 | Helped |
| Train the pseudo-label generator on COCO rather than transfer it from VOC (ablation) | Mask-RCNN instance segmentation · MS-COCO test | AP: 16.9 -> 22.2 | Helped |
| Use retrieval pseudo-labels rather than DenseCRF pseudo-labels (ablation) | BAP pseudo-labels on MS-COCO train (COCO-to-COCO) · MS-COCO train | AP: 17.2 -> 17.2 | Mixed |
Closest papers by meaning
- [Re] Background-Aware Pooling and Noise-Aware Loss for Weakly-Supervised Semantic Segmentation (NeurIPS 2022 · similarity 0.92)
- Combining Bottom-Up, Top-Down, and Smoothness Cues for Weakly Supervised Image Segmentation (CVPR 2017 · similarity 0.91)
- Simple Does It: Weakly Supervised Instance and Semantic Segmentation (CVPR 2017 · similarity 0.90)
- Weakly Supervised Semantic Segmentation Using Superpixel Pooling Network (AAAI 2017 · similarity 0.90)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.90)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.89)
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.