Research map / Point clouds / Shape and surface completion
MoGen: Detailed Neuronal Morphology Generation via Point Cloud Flow Matching
Franz Rieger, Jan-Matthis Lueckmann, Viren Jain, Michal Januszewski
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 k-NN relative-coordinate context injection (ablation) | MoGen · mixed mouse set | MMD: 4.27 -> 3.54 | Improved |
| add k-NN context injection (ablation) | MoGen · mixed mouse set | Loss (t in [0.8,0.9]): 0.712 -> 0.709 | Improved |
| double model width (ablation) | MoGen · mixed mouse set | MMD: 6.97 -> 3.54 | Improved |
| double model depth (ablation) | MoGen · mixed mouse set | MMD: 9.94 -> 3.54 | Improved |
| use modified cosine timestep schedule instead of linear (ablation) | MoGen · mixed mouse set | MMD: 5.12 -> 3.54 | Improved |
| train 1M steps instead of 500k | MoGen · mixed mouse set | MMD: 3.54 -> 3.08 | Improved |
| MoGen vs full transformer baseline | MoGen · 2048-point set | MMD (Total): 5.78 -> 4.94 | Improved |
| co-train SHAPE with 10% MoGen negatives | PATHFINDER SHAPE classifier · axon reconstruction | errors per mm: 0.7947 -> 0.7595 | Improved |
| replace 10% of negatives with synthetic (ablation) | SHAPE classifier · co-training | F1 (%): 93.91 -> 94.64 | Improved |
| replace 50% of data with synthetic (ablation) | SHAPE classifier · co-training | F1 (%): 94.64 -> 93.65 | Hurt |
| train on 100% synthetic (ablation) | SHAPE classifier · co-training | F1 (%): 93.91 -> 66.70 | Hurt |
| increase classifier-free guidance scale from 0.0 to 1.0 (ablation) | MoGen conditional generation · conditioned samples | faithfulness / diversity rank: faithfulness 2.20 -> 1.20; diversity 3.35 -> 4.10 | Mixed |
Closest papers by meaning
- MorphGrower: A Synchronized Layer-by-layer Growing Approach for Plausible Neuronal Morphology Generation (ICML 2024 · similarity 0.88)
- Learning Multimodal Volumetric Features for Large-Scale Neuron Tracing (AAAI 2024 · similarity 0.86)
- Cross-Dimension Affinity Distillation for 3D EM Neuron Segmentation (CVPR 2024 · similarity 0.85)
- FISBe: A Real-World Benchmark Dataset for Instance Segmentation of Long-Range Thin Filamentous Structures (CVPR 2024 · similarity 0.85)
- Efficient Connectivity-Preserving Instance Segmentation with Supervoxel-Based Loss Function (AAAI 2025 · similarity 0.84)
- SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation (NeurIPS 2026 · similarity 0.83)
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.
