Research map / Detection and segmentation / CNN features
Large Scale Image Segmentation with Structured Loss Based Deep Learning for Connectome Reconstruction
Jan Funke, Fabian Tschopp, William Grisaitis, Arlo Sheridan, Chandan Singh, Stephan Saalfeld, Srinivas C. Turaga
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 constrained MALIS training loss to the U-Net (ablation) | 3D U-Net + percentile agglomeration · FIB-25 (whole test volume) | VOI sum: 1.520 -> 1.071 | Improved |
| add constrained MALIS training loss to the U-Net (ablation) | 3D U-Net + percentile agglomeration · FIB-25 (synaptic sites) | VOI sum: 2.914 -> 2.151 | Improved |
| add constrained MALIS training loss to the U-Net (ablation) | 3D U-Net + percentile agglomeration · CREMI (test, leaderboard) | CREMI score: 0.793 -> 0.289 | Improved |
| add constrained MALIS training loss to the U-Net (ablation) | 3D U-Net + percentile agglomeration · SEGEM (skeletons) | IED total (um): 1.377 -> 4.839 | Improved |
| use U-NET MALA instead of FlyEM/CELIS+MC agglomeration pipelines | U-NET MALA · FIB-25 (whole test volume) | VOI sum: 1.266 -> 1.071 | Improved |
| use U-NET MALA instead of best leaderboard method LMC | U-NET MALA · CREMI (test, leaderboard) | CREMI score: 0.398 -> 0.289 | Improved |
| use 75% affinity quantile merge instead of 15% quantile (ablation) | U-NET MALA agglomeration · CREMI (training data, merge functions) | CREMI score: 0.212 -> 0.146 | Improved |
| use 50% quantile merge instead of mean affinity (ablation) | U-NET MALA agglomeration · FIB-25 (training data, merge functions) | VOI sum: 1.418 -> 1.350 | Improved |
| use 75% quantile merge instead of mean affinity (ablation) | U-NET MALA agglomeration · CREMI (training data, merge functions) | CREMI score: 0.149 -> 0.146 | Improved |
| use quantile-based agglomeration instead of Zlateski agglomeration (ablation) | U-NET MALA · CREMI (training data, merge functions) | CREMI score: 0.364 -> 0.146 | Improved |
| use quantile-based agglomeration instead of Zlateski agglomeration (ablation) | U-NET MALA · FIB-25 (training data, merge functions) | VOI sum: 2.071 -> 1.350 | Improved |
| discretize edge scores into 256 bins (ablation) | percentile agglomeration · EM agglomeration accuracy | accuracy | No clear effect |
Closest papers by meaning
- Efficient Neuron Segmentation in Electron Microscopy by Affinity-Guided Queries (ICLR 2025 · similarity 0.92)
- Cross-Dimension Affinity Distillation for 3D EM Neuron Segmentation (CVPR 2024 · similarity 0.91)
- Learning Multimodal Volumetric Features for Large-Scale Neuron Tracing (AAAI 2024 · similarity 0.89)
- Efficient Connectivity-Preserving Instance Segmentation with Supervoxel-Based Loss Function (AAAI 2025 · similarity 0.88)
- $\epsilon$-Seg: Sparsely Supervised Semantic Segmentation of Microscopy Data (NeurIPS 2025 · similarity 0.86)
- Electron Microscopy Images as Set of Fragments for Mitochondrial Segmentation (AAAI 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.
