Research map / Architectures and efficiency / Neural architecture search
DiNTS: Differentiable Neural Network Topology Search for 3D Medical Image Segmentation
Check 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 |
|---|---|---|---|
| Use DiNTS instead of nnUNet | MSD overall test · MSD overall test | DSC: 77.89 -> 77.93 | Helped |
| Use DiNTS instead of nnUNet | MSD overall test · MSD overall test | NSD: 88.09 -> 88.68 | Helped |
| Reduce memory budget from σ=0.8 to σ=0.5 (ablation) | DiNTS on Pancreas · Pancreas 5-fold cross-validation | Avg. DSC: 66.29 -> 66.23 | No clear effect |
| Reduce memory budget from σ=0.8 to σ=0.5 (ablation) | DiNTS on Pancreas · Pancreas 5-fold cross-validation | GPU memory: 13018 -> 10110 | Helped |
| Reduce memory budget from σ=0.8 to σ=0.2 (ablation) | DiNTS on Pancreas · Pancreas 5-fold cross-validation | Avg. DSC: 66.29 -> 63.00 | Hurt |
| Search architecture on Brain rather than Pancreas (ablation) | DiNTS on Brain · Brain 5-fold cross-validation | Avg. DSC: 72.61 -> 72.97 | Helped |
| Search architecture on Liver rather than Pancreas (ablation) | DiNTS on Liver · Liver 5-fold cross-validation | Avg. DSC: 75.99 -> 76.44 | Helped |
| Search architecture on Lung rather than Pancreas (ablation) | DiNTS on Lung · Lung 5-fold cross-validation | DSC1: 68.90 -> 69.30 | Helped |
| Remove topology loss during search (ablation) | DiNTS · Architecture search | discretization gap G | Hurt |
| Search with DiNTS rather than C2FNAS | 3D medical image segmentation NAS · MSD Pancreas search | search time: 333 GPU days -> 5.8 GPU days | Helped |
Closest papers by meaning
- C2FNAS: Coarse-to-Fine Neural Architecture Search for 3D Medical Image Segmentation (CVPR 2020 · similarity 0.91)
- HyperSegNAS: Bridging One-Shot Neural Architecture Search With 3D Medical Image Segmentation Using HyperNet (CVPR 2022 · similarity 0.89)
- DOTS: Decoupling Operation and Topology in Differentiable Architecture Search (CVPR 2021 · similarity 0.86)
- Upping the Game: How 2D U-Net Skip Connections Flip 3D Segmentation (NeurIPS 2024 · similarity 0.85)
- E2ENet: Dynamic Sparse Feature Fusion for Accurate and Efficient 3D Medical Image Segmentation (NeurIPS 2024 · similarity 0.85)
- EffiDec3D: An Optimized Decoder for High-Performance and Efficient 3D Medical Image Segmentation (CVPR 2025 · similarity 0.84)
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.