Research map / Detection and segmentation / Weakly supervised detection and segmentation
Semi-Supervised Semantic Image Segmentation With Self-Correcting Networks
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 |
|---|---|---|---|
| Add convolutional self-correction to refine weak-set labels (ablation) | DeepLabv3+ primary segmentation model · PASCAL VOC 2012 validation, F=1464 and W=9118 | mIOU: 80.34 -> 82.33 | Helped |
| Add linear self-correction to combine ancillary and primary predictions (ablation) | DeepLabv3+ primary segmentation model · PASCAL VOC 2012 validation, F=1464 and W=9118 | mIOU: 80.34 -> 81.35 | Helped |
| Add convolutional self-correction to refine weak-set labels (ablation) | DeepLabv3+ primary segmentation model · Cityscapes validation, F=914 and W=2061 | mIOU: 75.44 -> 79.46 | Helped |
| Add linear self-correction to combine ancillary and primary predictions (ablation) | DeepLabv3+ primary segmentation model · Cityscapes validation, F=914 and W=2061 | mIOU: 75.44 -> 76.22 | Helped |
| Use convolutional rather than linear self-correction (ablation) | Self-correcting DeepLabv3+ · Cityscapes validation, F=914 and W=2061 | mIOU: 76.22 -> 79.46 | Helped |
| Use convolutional rather than linear self-correction (ablation) | Self-correcting DeepLabv3+ · PASCAL VOC 2012 validation, F=1464 and W=9118 | mIOU: 81.35 -> 82.33 | Helped |
| Use convolutional rather than linear self-correction (ablation) | Self-correcting DeepLabv3+ · PASCAL VOC 2012 validation, F=200 and W=10382 | mIOU: 79.43 -> 78.29 | Hurt |
| Use convolutional rather than linear self-correction (ablation) | Self-correcting DeepLabv3+ · Cityscapes validation, F=200 and W=2775 | mIOU: 73.56 -> 69.38 | Hurt |
| Use convolutional rather than linear self-correction (ablation) | Self-correcting DeepLabv3+ · Cityscapes validation, F=450 and W=2525 | mIOU: 75.24 -> 77.16 | Helped |
| Replace fully supervised training with semi-supervised training using 200 fully labeled images and box-labeled weak images | Linear self-correcting DeepLabv3+ · PASCAL VOC 2012 validation | mIOU | No clear effect |
| Train convolutional self-correction to refine ground-truth labels rather than use them unchanged (ablation) | DeepLabv3+ on PASCAL VOC 2012 original training set · PASCAL VOC 2012 validation | mAP: 68.8% -> 76.88% | Helped |
| Use convolutional self-correction on top of bounding boxes (ablation) | DeepLabv3+ · PASCAL VOC 2012 validation | mAP: 68.8% -> 75.97% | Helped |
Closest papers by meaning
- Single-Stage Semantic Segmentation From Image Labels (CVPR 2020 · similarity 0.91)
- Three Ways To Improve Semantic Segmentation With Self-Supervised Depth Estimation (CVPR 2021 · similarity 0.91)
- Simple Does It: Weakly Supervised Instance and Semantic Segmentation (CVPR 2017 · similarity 0.91)
- GuidedMix-Net: Semi-supervised Semantic Segmentation by Using Labeled Images as Reference (AAAI 2022 · similarity 0.90)
- Combining Bottom-Up, Top-Down, and Smoothness Cues for Weakly Supervised Image Segmentation (CVPR 2017 · similarity 0.90)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.90)
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.