Research map / Detection and segmentation / CNN features
Topology-Aware Segmentation Using Discrete Morse Theory
Xiaoling Hu, Yusu Wang, Li Fuxin, Dimitris Samaras, Chao Chen
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 DMT-loss instead of TopoLoss | U-Net segmentation on CREMI 2D · CREMI 2D | Betti Error: 1.113 -> 0.982 | Improved |
| Use DMT-loss instead of TopoLoss | segmentation networks · ISBI13 2D | Betti Error: 1.253 ± 0.172 -> 1.102 ± 0.203 | Improved |
| Use DMT-loss instead of TopoLoss | segmentation networks · CrackTree | Betti Error: 0.672 ± 0.176 -> 0.518 ± 0.189 | Improved |
| Use DMT-loss instead of TopoLoss | segmentation networks · Road | Betti Error: 1.275 ± 0.192 -> 0.995 ± 0.301 | Improved |
| Use DMT-loss instead of TopoLoss | 3D segmentation networks · ISBI13 3D | Betti Error: 0.972 ± 0.245 -> 0.812 ± 0.134 | Improved |
| Use DMT-loss instead of TopoLoss | 3D segmentation networks · CREMI 3D | Betti Error: 1.076 ± 0.206 -> 0.726 ± 0.187 | Improved |
| Use DMT-loss instead of TopoLoss | 3D segmentation networks · 3Dircadb | Betti Error: 2.245 ± 0.255 -> 1.415 ± 0.305 | Improved |
| Set persistence threshold epsilon to 0.1 rather than 0.2 (ablation) | DMT-loss · CREMI 2D | Betti Error: 0.982 -> 1.206 | Hurt |
| Set persistence threshold epsilon to 0.3 rather than 0.2 (ablation) | DMT-loss · CREMI 2D | Betti Error: 0.982 -> 2.105 | Hurt |
| Set balanced-loss weight beta to 2 rather than 3 (ablation) | DMT-loss · CREMI 2D | Betti Error: 0.982 -> 1.074 | Hurt |
| Set balanced-loss weight beta to 4 rather than 3 (ablation) | DMT-loss · CREMI 2D | Betti Error: 0.982 -> 1.181 | Hurt |
| Use DMT-selected critical structures rather than generic reweighting of false-positive and false-negative pixels | cross-entropy segmentation loss · CREMI 2D | Betti Error: 2.753 vs. 0.982 | Improved |
Closest papers by meaning
- Structure-Aware Image Segmentation with Homotopy Warping (NeurIPS 2022 · similarity 0.92)
- Topology-Aware Uncertainty for Image Segmentation (NeurIPS 2023 · similarity 0.91)
- Learning Probabilistic Topological Representations Using Discrete Morse Theory (ICLR 2023 · similarity 0.91)
- Topograph: An Efficient Graph-Based Framework for Strictly Topology Preserving Image Segmentation (ICLR 2025 · similarity 0.90)
- A Topological Loss Function for Deep-Learning Based Image Segmentation Using Persistent Homology (TPAMI 2020 · similarity 0.90)
- Scale-Free Image Keypoints Using Differentiable Persistent Homology (ICML 2024 · 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.
