Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-Supervised Semantic Segmentation under Label Noise via Diverse Learning Groups
Peixia Li, Pulak Purkait, Thalaiyasingam Ajanthan, Majid Abdolshah, Ravi Garg, Hisham Husain, Chenchen Xu, Stephen Gould, Wanli Ouyang, Anton van den Hengel
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 diverse learning groups and filter module to semi-supervised segmentation under 9% PL noise (ablation) | PS-MT · Pascal VOC 2012, 1/8 labelled, 9% PL | mIoU (%): 65.1 -> 74.0 | Improved |
| add filter module to the two-LG framework (ablation) | Ours · Pascal VOC 2012, 1/8 labelled, 9% PL | mIoU (%): 69.9 -> 74.0 | Improved |
| add filter module to single-LG PS-MT (ablation) | PS-MT (R50) · Pascal VOC 2012, 1/8 labelled, 9% PL | mIoU (%): 65.1 -> 67.6 | Improved |
| use teacher from the other learning group for noise detection instead of own group (ablation) | Ours · Pascal VOC 2012, 1/8 labelled, 9% PL | mIoU (%): 72.4 -> 74.0 | Improved |
| use different architectures (R50 and PVTv2-B2) in the two learning groups instead of same | Ours · Pascal VOC 2012 | mIoU | Improved |
| use CE+KL score instead of CE or KL alone for noise detection | filter module · noise detection | F1 of noise detection | Improved |
| use proposed filter module instead of GMM for noise detection | filter module · noise detection | detected noise proportion | Improved |
| apply method to clean-labelled semi-supervised segmentation (0% noise) | Ours versus PS-MT · Pascal VOC 2012, 1/2 labelled | mIoU (%): 77.88 -> 79.54 | Improved |
| apply method versus PS-MT under SCP noise at 1/2 labels | Ours versus PS-MT · Pascal VOC 2012, 9% SCP | mIoU (%): 53.69 -> 75.95 | Improved |
| apply framework to weakly supervised segmentation with PVTv2 baseline | PS-MT (PVTv2-B2) · Pascal VOC 2012 weakly supervised | mIoU (%): 66.7 -> 71.7 | Improved |
Closest papers by meaning
- Semi-Supervised Semantic Segmentation With High- and Low-Level Consistency (TPAMI 2019 · similarity 0.93)
- Semi-Supervised Semantic Segmentation Using Unreliable Pseudo-Labels (CVPR 2022 · similarity 0.92)
- Semi-supervised Semantic Segmentation via Strong-weak Dual-branch Network (ECCV 2020 · similarity 0.91)
- Semi-Supervised Semantic Segmentation With Pixel-Level Contrastive Learning From a Class-Wise Memory Bank (ICCV 2021 · similarity 0.91)
- Semi-Supervised Segmentation based on Error-Correcting Supervision (ECCV 2020 · similarity 0.91)
- AllSpark: Reborn Labeled Features from Unlabeled in Transformer for Semi-Supervised Semantic Segmentation (CVPR 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.
