Research map / Label-efficient and robust learning / Semi-supervised learning
DualFete: Revisiting Teacher-Student Interactions from a Feedback Perspective for Semi-supervised Medical Image Segmentation
Le Yi, Wei Huang, Lei Zhang, Kefu Zhao, Yan Wang, Zizhou Wang
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 dual-teacher feedback (agreement+disagreement) to cross-supervised teachers, full student (ablation) | DualFete student · LA 20% labels | Dice (%): 91.10 -> 91.89 | Improved |
| add feedback loss to cross-supervised dual teacher (ablation) | DualFete student · Pancreas 20% labels | Dice (%): 82.17 -> 83.49 | Improved |
| add cross-supervision between two teachers (ablation) | teacher-student with labeled loss · LA 20% labels | Dice (%): 88.55 -> 90.30 | Improved |
| add single-teacher feedback to baseline teacher-student (ablation) | teacher-student · Pancreas 20% labels | Dice (%): 77.18 -> 79.27 | Improved |
| go from single-teacher to dual-teacher feedback with matched attributors and receivers (ablation) | DualFete · Pancreas 20% labels | Dice (%): 79.27 -> 81.12 | Improved |
| mismatch feedback attributors and receivers (ablation) | dual-teacher DualFete · LA 20% labels | Dice (%): 90.89 -> 87.69 | Hurt |
| use strong-augmentation likelihood in feedback loss (ablation) | dual-teacher DualFete · LA 20% labels | Dice (%): 90.89 -> 90.14 | Hurt |
| use feedback loss alone without cross-supervision (ablation) | dual-teacher DualFete · Pancreas 20% labels | Dice under strong-augmentation perturbation: 78.42 -> 57.40 (teacher phi) | Hurt |
| fine-tune the student on labeled data (w.ft.) (ablation) | DualFete · LA 5% and 10% labels | Dice (%): 90.35 -> 90.22 (LA 5%); Pancreas 20%: 83.49 -> 83.85 | Mixed |
| use DualFete instead of AD-MT | DualFete · Pancreas 10% labels | Dice (%): 80.21 -> 81.99 | Improved |
| use DualFete instead of TraCoCo | DualFete · BraTS 20% labels | Dice (%): 86.69 -> 85.83 | Hurt |
| set confidence threshold of cross-supervised loss to 0.7 (ablation) | DualFete · LA and Pancreas | Dice (%) | Improved |
Closest papers by meaning
- Alternate Diverse Teaching for Semi-supervised Medical Image Segmentation (ECCV 2024 · similarity 0.90)
- PMT: Progressive Mean Teacher via Exploring Temporal Consistency for Semi-Supervised Medical Image Segmentation (ECCV 2024 · similarity 0.88)
- Medical Image Segmentation with Minimal Labeling Effort: How Far Can We Push the Limits? (AAAI 2026 · similarity 0.87)
- ATSO: Asynchronous Teacher-Student Optimization for Semi-Supervised Image Segmentation (CVPR 2021 · similarity 0.87)
- Steady Progress Beats Stagnation: Mutual Aid of Foundation and Conventional Models in Mixed Domain Semi-Supervised Medical Image Segmentation (CVPR 2025 · similarity 0.86)
- GapMatch: Bridging Instance and Model Perturbations for Enhanced Semi-Supervised Medical Image Segmentation (AAAI 2025 · similarity 0.85)
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.
