Research map / Detection and segmentation / CNN features
Topologically Faithful Image Segmentation via Induced Matching of Persistence Barcodes
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 |
|---|---|---|---|
| add Betti matching loss on top of Dice loss | segmentation network · CREMI | Betti matching error τ^err: 74.82 -> 64.9001 | Helped |
| add Betti matching loss on top of Dice loss | segmentation network · Roads | Betti matching error τ^err: 58.90 -> 41.50 | Helped |
| add Betti matching loss on top of Dice loss | segmentation network · synMnist | Betti matching error τ^err: 1.849 -> 1.140 | Helped |
| add Betti matching loss on top of Dice loss | segmentation network · Buildings | Betti matching error τ^err: Hu et al. 278.30 -> 244.58 | Helped |
| add Betti matching loss on top of Dice loss | segmentation network · Colon | Betti matching error τ^err: Hu et al. 17.25 vs Ours 16.00 | Mixed |
| increase alpha for Betti matching loss (ablation) | Roads (Betti matching loss) · Roads | Betti matching error τ^err: 53.958 at alpha 0.0005 -> 41.50 at alpha 0.5 | Helped |
| increase alpha for Betti matching loss (ablation) | synMnist (Betti matching loss) · synMnist | Dice: 0.871 at alpha 0.005 -> 0.796 at alpha 0.5 | Hurt |
| match both foreground and background levels (bothlevel) instead of foreground only (superlevel) (ablation) | Betti matching loss · Elegans | Betti matching error τ^err: 2.15 -> 1.70 | Helped |
| match both foreground and background levels (bothlevel) instead of foreground only (superlevel) (ablation) | Betti matching loss · CREMI | Betti matching error τ^err: 59.20 (superlevel) vs 60.48 (bothlevel) | No clear effect |
| add image boundary (relative Betti matching) (ablation) | Betti matching loss · Roads | Betti matching error τ^err: 41.500 relative vs 41.417 non-relative at alpha 0.5 | No clear effect |
| apply Betti matching loss to Hu et al. method (dim-0,1 matching instead of dim-1 only) (ablation) | Hu et al. method · Roads | Betti matching error τ^err: dim-1 alpha 0.5: 58.583 vs dim-0,1 alpha 0.5: 44.708 | Mixed |
| train Betti matching loss from scratch vs pretrained (ablation) | Betti matching loss · CREMI | Betti matching error τ^err: qualitative claim only | No clear effect |
Closest papers by meaning
- Topograph: An Efficient Graph-Based Framework for Strictly Topology Preserving Image Segmentation (ICLR 2025 · similarity 0.91)
- Topology-Preserving Deep Image Segmentation (NeurIPS 2019 · similarity 0.90)
- Towards Scalable Topological Regularizers (ICLR 2025 · similarity 0.86)
- MATCH: Multi-faceted Adaptive Topo-Consistency for Semi-Supervised Histopathology Segmentation (NeurIPS 2025 · similarity 0.86)
- TopoDiffusionNet: A Topology-aware Diffusion Model (ICLR 2025 · similarity 0.85)
- ToMAToMP: Robust and Multi-Parameter Topological Clustering (NeurIPS 2026 · 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.