Research map / Label-efficient and robust learning / Semi-supervised learning
ConformalSAM: Unlocking the Potential of Foundational Segmentation Models in Semi-Supervised Semantic Segmentation with Conformal Prediction
Danhui Chen, Ziquan Liu, Chuxi Yang, Dan Wang, Yan Yan, Yi Xu, Xiangyang Ji
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 |
|---|---|---|---|
| use raw SEEM masks as pseudo-labels for unlabeled data | SegFormer-B5 label-only · PASCAL VOC 2012 original, 1/16 (92) | mIoU: 50.65 -> 42.00 | Hurt |
| use raw SEEM masks as pseudo-labels for unlabeled data | SegFormer-B5 label-only · ADE20K, 1/16 (1263) | mIoU: 28.76 -> 21.45 | Hurt |
| use raw SEEM masks as pseudo-labels (smallest ADE20K split) | SegFormer-B5 label-only · ADE20K, 1/64 (316) | mIoU: 18.83 -> 20.53 | Improved |
| add conformal prediction calibration to SEEM pseudo-labels (ablation) | ConformalSAM · PASCAL VOC 2012 original | mIoU: 42.00 -> 78.09 | Improved |
| add self-reliance training in stage II (ablation) | ConformalSAM · PASCAL VOC 2012 original | mIoU: 78.09 -> 81.21 | Improved |
| add self-reliance training without CP (ablation) | SEEM pseudo-labels · PASCAL VOC 2012 original | mIoU: 42.00 -> 65.94 | Improved |
| use mis-coverage rate alpha 0.05 instead of 0.1 (ablation) | pixel-wise CP · PASCAL VOC 2012 original | mIoU: 74.31 -> 78.09 | Improved |
| use mis-coverage rate alpha 0.05 instead of 0.01 (ablation) | pixel-wise CP · PASCAL VOC 2012 original | mIoU: 68.01 -> 78.09 | Improved |
| use pixel-wise CP instead of image-level CP (ablation) | ConformalSAM · PASCAL VOC 2012 original | mIoU: 75.54 -> 78.09 | Improved |
| add class-conditional label selection to vanilla CP (ablation) | ConformalSAM · PASCAL VOC 2012 original | mIoU: 34.11 mIoU average | Improved |
| start self-reliance training at epoch 60 instead of 10 (ablation) | ConformalSAM · PASCAL VOC 2012 original | mIoU | Improved |
| combine ConformalSAM with AllSpark instead of AllSpark alone | AllSpark · PASCAL VOC 2012 original | mIoU: 76.07 -> 80.69 | Improved |
Closest papers by meaning
- Learning from Future: A Novel Self-Training Framework for Semantic Segmentation (NeurIPS 2022 · similarity 0.89)
- Universal Semi-Supervised Semantic Segmentation (ICCV 2019 · similarity 0.89)
- Label-Efficient Few-Shot Semantic Segmentation with Unsupervised Meta-Training (AAAI 2024 · similarity 0.89)
- Single-Stage Semantic Segmentation From Image Labels (CVPR 2020 · similarity 0.88)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.88)
- SemiDAViL: Semi-supervised Domain Adaptation with Vision-Language Guidance for Semantic Segmentation (CVPR 2025 · 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.
