Research map / Detection and segmentation / CNN features
UCTransNet: Rethinking the Skip Connections in U-Net from a Channel-Wise Perspective with Transformer
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace plain U-Net skip connections with CCT and CCA (ablation) | U-Net · GlaS | Dice: 85.45 -> 89.84 | Helped |
| replace plain U-Net skip connections with CCT and CCA (ablation) | U-Net · MoNuSeg | Dice: 76.45 -> 79.87 | Helped |
| add CCT for multi-scale encoder feature fusion (ablation) | U-Net · GlaS | Dice: 85.45 -> 89.09 | Helped |
| add CCT for multi-scale encoder feature fusion (ablation) | U-Net · MoNuSeg | Dice: 76.45 -> 79.31 | Helped |
| add CCA for decoder-feature fusion (ablation) | U-Net · GlaS | Dice: 85.45 -> 87.09 | Helped |
| add CCA for decoder-feature fusion (ablation) | U-Net · MoNuSeg | Dice: 76.45 -> 76.84 | Helped |
| remove CCA (ablation) | UCTransNet · Synapse | Dice: 78.23 -> 78.99 | Hurt |
| remove CCA (ablation) | UCTransNet · Synapse | HD: 26.75 -> 30.29 | Hurt |
| pre-train U-Net before training UCTransNet | UCTransNet · Synapse | Dice: 78.23 -> 75.54 | Hurt |
| pre-train U-Net before training UCTransNet | UCTransNet · Synapse | HD: 26.75 -> 38.97 | Hurt |
| use channel-wise skip fusion instead of the original U-Net connections | U-Net · GlaS | Dice: 85.45 -> 90.18$\pm$0.71$^\ast$ | Helped |
| use channel-wise skip fusion instead of the original U-Net connections | U-Net · MoNuSeg | Dice: 76.45$\pm$2.62 -> 79.08$\pm$0.67 | Helped |
Closest papers by meaning
- Rethinking U-Net: Task-Adaptive Mixture of Skip Connections for Enhanced Medical Image Segmentation (AAAI 2025 · similarity 0.90)
- Upping the Game: How 2D U-Net Skip Connections Flip 3D Segmentation (NeurIPS 2024 · similarity 0.89)
- nnWNet: Rethinking the Use of Transformers in Biomedical Image Segmentation and Calling for a Unified Evaluation Benchmark (CVPR 2025 · similarity 0.88)
- FuseUNet: A Multi-Scale Feature Fusion Method for U-like Networks (ICML 2025 · similarity 0.86)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.85)
- Toward Real Ultra Image Segmentation: Leveraging Surrounding Context to Cultivate General Segmentation Model (NeurIPS 2024 · 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.