Research map / Label-efficient and robust learning / Semi-supervised learning
Graph-Constrained Contrastive Regularization for Semi-Weakly Volumetric Segmentation
Simon Reiß, Constantin Seibold, Alexander Freytag, Erik Rodner, Rainer Stiefelhagen
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 Con2R regularization instead of best competing semi-supervised method | 3D UNet · RETOUCH M(24) | mIoU: 48.2 vs 46.2 (FixMatch) | Improved |
| add Con2R to 3D UNet | 3D UNet · RETOUCH M(3) | mIoU: 12.0 -> 14.8 | Improved |
| add Con2R to 3D UNet | 3D UNet · RETOUCH M(12) | mIoU: 31.1 -> 38.6 | Improved |
| add Con2R to 3D UNet | 3D UNet · RETOUCH M(24) | mIoU: 43.8 -> 48.2 | Improved |
| add Con2R when all volumes are fully annotated | 3D UNet · RETOUCH M(full) | mIoU: 54.9 -> 54.6 | No clear effect |
| add Con2R to 3D UNet | 3D UNet · BraTS M(24) | mIoU (average): 38.8 -> 43.2 | Improved |
| use 16x16x16 receptive volume instead of 32x32x32 (ablation) | Con2R · RETOUCH M(24) validation | validation mIoU: 47.1 -> 49.1 | Improved |
| use 16x16x16 receptive volume instead of 160x160x160 (ablation) | Con2R · RETOUCH M(24) validation | validation mIoU: 46.5 -> 49.1 | Improved |
| increase number of sampled graph vertices from 216 to 1728 (ablation) | Con2R · RETOUCH M(24) validation | validation mIoU: 46.9 -> 49.1 | Improved |
| increase number of sampled graph vertices from 1000 to 1728 (ablation) | Con2R · RETOUCH M(24) validation | validation mIoU: 47.8 -> 49.1 | Improved |
| mix positional and semantic constraints with alpha 0.2 rather than only one (ablation) | Con2R · RETOUCH M(24) validation | validation mIoU | Improved |
| use Con2R vs 3D UNet on SRF class | 3D UNet · RETOUCH M(24) SRF | IoU: 73.0 -> 79.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FixMatch — FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (NeurIPS 2020) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Towards Generic Semi-Supervised Framework for Volumetric Medical Image Segmentation (NeurIPS 2023 · similarity 0.89)
- Weakly Supervised 3D Semantic Segmentation Using Cross-Image Consensus and Inter-Voxel Affinity Relations (ICCV 2021 · similarity 0.88)
- SGTC: Semantic-Guided Triplet Co-training for Sparsely Annotated Semi-Supervised Medical Image Segmentation (AAAI 2025 · similarity 0.86)
- Raptor: Scalable Train-Free Embeddings for 3D Medical Volumes Leveraging Pretrained 2D Foundation Models (ICML 2025 · similarity 0.86)
- Neighbor Does Matter: Density-Aware Contrastive Learning for Medical Semi-supervised Segmentation (AAAI 2025 · similarity 0.86)
- Learning General-purpose Biomedical Volume Representations using Randomized Synthesis (ICLR 2025 · similarity 0.86)
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.
