Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-supervised Semantic Segmentation via Strong-weak Dual-branch Network
Wenfeng Luo, Meng Yang
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 9k weak annotations to 1.4k strong ones in a single-branch network (ablation) | single-branch VGG16 · PASCAL VOC val | mIoU: 68.9 -> 62.8 | Hurt |
| use a strong-weak dual-branch network instead of single branch with same data (ablation) | VGG16 segmentation net · PASCAL VOC val | mIoU: 62.8 -> 72.2 | Improved |
| add oversampling of strong data to dual-branch network (ablation) | dual-branch VGG16 · PASCAL VOC val | mIoU: 66.4 -> 72.2 | Improved |
| oversample strong annotations in single-branch network (ablation) | single-branch VGG16 · PASCAL VOC val | mIoU: 62.8 -> 65.9 | Mixed |
| use 10k strong annotations in the weak branch instead of 1.4k strong + 9k weak (ablation) | dual-branch VGG16 · PASCAL VOC val | mIoU: 72.2 -> 73.9 | Improved |
| ensemble separate strong and weak networks with lambda 0.4 (ablation) | strong-only network · PASCAL VOC val | mIoU: 68.9 -> 70.8 | Improved |
| add weak branch (dual-branch) on top of 1.4k strong only (ablation) | strong-only single branch · PASCAL VOC val | mIoU: 68.9 -> 72.2 | Improved |
| dual-branch with weak labels vs single-branch trained on 10k strong | dual-branch VGG16 · PASCAL VOC val | mIoU: 71.4 -> 72.2 | Improved |
| add 98k weak annotations to 20k strong in single-branch network (ablation) | single-branch · COCO val | mean IoU: 46.1 -> 33.4 | Hurt |
| use dual-branch network with 98k weak annotations (ablation) | dual-branch · COCO val | mean IoU: 33.4 -> 47.6 | Improved |
| switch backbone from VGG16 to ResNet101 | dual-branch · PASCAL VOC test | mIoU: 72.3 -> 77.1 | Improved |
Closest papers by meaning
- Semi-Supervised Semantic Segmentation With High- and Low-Level Consistency (TPAMI 2019 · similarity 0.93)
- Single-Stage Semantic Segmentation From Image Labels (CVPR 2020 · similarity 0.92)
- Reliability Does Matter: An End-to-End Weakly Supervised Semantic Segmentation Approach (AAAI 2020 · similarity 0.91)
- Universal Weakly Supervised Segmentation by Pixel-to-Segment Contrastive Learning (ICLR 2021 · similarity 0.91)
- Semi-Supervised Semantic Image Segmentation With Self-Correcting Networks (CVPR 2020 · similarity 0.91)
- W2P: Switching from Weak Supervision to Partial Supervision for 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.
