Research map / Label-efficient and robust learning / Self-supervised and contrastive
NeuroSeg Meets DINOv3: Transferring 2D Self-Supervised Visual Priors to 3D Neuron Segmentation via DINOv3 Initialization
Yik San Cheng, Runkai Zhao, Weidong Cai
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 center inflation instead of average inflation (ablation) | NeurINO-T · Drosophila | F1 (%): 49.64 -> 50.06 | Improved |
| use center inflation instead of average inflation (ablation) | NeurINO-T · Drosophila SmartTracing | ESA: 1.71 -> 1.62 | Improved |
| add topology-aware skeleton loss (TASL) (ablation) | NeurINO-T · Drosophila | F1 (%): 50.59 -> 50.06 | Hurt |
| add topology-aware skeleton loss (TASL) (ablation) | NeurINO-T · Drosophila | HD95: 2.99 -> 3.07 | Hurt |
| add topology-aware skeleton loss (TASL) (ablation) | NeurINO-T · Drosophila SmartTracing | ESA: 1.68 -> 1.62 | Improved |
| add topology-aware skeleton loss (TASL) (ablation) | NeurINO-T · Drosophila NeuTube | ESA: 1.94 -> 1.75 | Improved |
| fine-tune the pretrained DINOv3 encoder instead of freezing it (ablation) | NeurINO-T · Drosophila | F1 (%): 47.22 -> 50.06 | Improved |
| fine-tune the pretrained DINOv3 encoder instead of freezing it (ablation) | NeurINO-T · Drosophila | HD95: 5.89 -> 3.07 | Improved |
| initialize from DINOv3 pretrained weights instead of from scratch (ablation) | NeurINO-T · Drosophila | F1 (%): 48.56 -> 50.06 | Improved |
| use 2x first downsampling instead of 4x in DINOv3 (ablation) | NeurINO-T · Drosophila | F1 (%): 47.08 -> 50.06 | Improved |
| add TASL to nnUNet (ablation) | nnUNet · Drosophila | F1 (%): 47.20 -> 47.09 | No clear effect |
| add TASL to nnUNet (ablation) | nnUNet · Drosophila SmartTracing | ESA: 1.67 -> 1.65 | Improved |
Closest papers by meaning
- Cross-Dimension Affinity Distillation for 3D EM Neuron Segmentation (CVPR 2024 · similarity 0.87)
- Learning Multimodal Volumetric Features for Large-Scale Neuron Tracing (AAAI 2024 · similarity 0.86)
- VISTA3D: A Unified Segmentation Foundation Model For 3D Medical Imaging (CVPR 2025 · similarity 0.85)
- DinoComplete: 3D Shape Completion with Distilled Semantic Priors and State Space Models (NeurIPS 2026 · similarity 0.85)
- Efficient Connectivity-Preserving Instance Segmentation with Supervoxel-Based Loss Function (AAAI 2025 · similarity 0.85)
- Neuron Platonic Intrinsic Representation From Dynamics Using Contrastive Learning (ICLR 2025 · 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.
