Research map / Detection and segmentation / CNN features
Towards High-Quality Image Segmentation: Improving Topology Accuracy by Penalizing Neighbor Pixels
Juan Miguel Valverde, Dim P. Papadopoulos, Rasmus Larsen, Anders Bjorholm Dahl
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 SCNP to Cross Entropy Dice loss | CEDice · FIVES | Betti error b0: 10.82 -> 6.94 | Improved |
| Optimize SCNP-penalized logits with Cross Entropy (ablation) | CE · FIVES | Betti error b0: 11.93 -> 7.53 | Improved |
| Optimize SCNP-penalized logits with Dice loss (ablation) | Dice · FIVES | Betti error b0: 12.03 -> 7.88 | Improved |
| Optimize SCNP-penalized logits with clDice (ablation) | clDice (i=24) · FIVES | Betti error b0: 36.55 -> 5.44 | Improved |
| Optimize SCNP-penalized logits with SkelRecall (ablation) | SkelRecall · FIVES | Betti error b0: 12.45 -> 5.07 | Improved |
| Optimize SCNP-penalized logits with TopoLoss (ablation) | TopoLoss · FIVES | Betti error b0: 8.09 -> 6.36 | Improved |
| Optimize SCNP-penalized logits with RWLoss (ablation) | RWLoss · FIVES | Betti error b0: 15.99 -> 8.64 | Improved |
| Optimize SCNP-penalized logits with Focal loss (ablation) | Focal · FIVES | Betti error b0: 16.08 -> 7.75 | Improved |
| Increase SCNP window size from 3 to 9 (ablation) | CEDice + SCNP · FIVES 2048 px | Betti error b0: 6.94 -> 4.79 | Improved |
| Increase SCNP window size from 9 to 11 (ablation) | CEDice + SCNP · FIVES 2048 px | Betti error b0: 4.79 -> 7.80 | Hurt |
| Add SCNP to CEDice on DeepGlobeRoads | CEDice · DeepGlobeRoads | Betti error b0 | Mixed |
| Add SCNP to CEDice on CirrMRI600 | CEDice · CirrMRI600 (non-tubular) | Betti error b0 | Improved |
Closest papers by meaning
- Topology-Preserving Deep Image Segmentation (NeurIPS 2019 · similarity 0.90)
- Topograph: An Efficient Graph-Based Framework for Strictly Topology Preserving Image Segmentation (ICLR 2025 · similarity 0.89)
- Structure-Aware Image Segmentation with Homotopy Warping (NeurIPS 2022 · similarity 0.89)
- Topology-Aware Segmentation Using Discrete Morse Theory (ICLR 2021 · similarity 0.89)
- SegGraph: Leveraging Graphs of SAM Segments for Few-Shot 3D Part Segmentation (NeurIPS 2025 · similarity 0.85)
- Efficient Connectivity-Preserving Instance Segmentation with Supervoxel-Based Loss Function (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.
