Research map / Label-efficient and robust learning / Self-supervised and contrastive
ASCENT: Annotation-free Self-supervised Contrastive Embeddings for 3D Neuron Tracking in Fluorescence Microscopy
Haejun Han, Hang Lu
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 |
|---|---|---|---|
| replace single MIP with iterative volume compression (ablation) | VCE-Net · NeRVE (100 frames) | plateau accuracy: 83.30 -> 95.04 | Improved |
| use volume compression instead of full 3D convolution (ablation) | VCE-Net · NeRVE (100 frames) | plateau accuracy: 57.80 -> 95.04 | Improved |
| add local patch pathway (ablation) | VCE-Net · NeRVE (100 frames) | plateau accuracy: 92.98 -> 95.04 | Improved |
| use ResNet-34 instead of ResNet-18 backbone (ablation) | VCE-Net · NeRVE (100 frames) | plateau accuracy: 95.04 -> 95.09 | No clear effect |
| use ResNet-50 instead of ResNet-18 backbone (ablation) | VCE-Net · NeRVE (100 frames) | plateau accuracy: 95.04 -> 95.02 | No clear effect |
| train on 200 instead of 100 regularly sampled frames (ablation) | ASCENT · NeRVE | plateau accuracy: 95.04% vs. 95.20% | Improved |
| use uniformly sampled 100 frames instead of first 100 frames (ablation) | ASCENT · NeRVE | plateau accuracy: 94.44% vs. 95.04% | Improved |
| train on only 12 frames (ablation) | ASCENT · NeRVE | plateau accuracy: around 93% | Mixed |
| add data augmentations (all five transforms vs none) (ablation) | ASCENT · NeRVE (50 frames) | plateau accuracy: 44.5% -> 94.4% | Improved |
| add position jitter augmentation (ablation) | ASCENT · NeRVE (50 frames) | plateau accuracy: -7.1% when omitted; +45.4% alone | Improved |
| add elastic deformation augmentation (ablation) | ASCENT · NeRVE (50 frames) | plateau accuracy: +0.1% when removed | No clear effect |
| train on NeRVE rather than in-house data, test on NeRVE (ablation) | ASCENT · NeRVE test | tracking accuracy after epoch 100: 71.73% -> 95.07% | Improved |
Closest papers by meaning
- Cross-Dimension Affinity Distillation for 3D EM Neuron Segmentation (CVPR 2024 · similarity 0.86)
- Learning Multimodal Volumetric Features for Large-Scale Neuron Tracing (AAAI 2024 · similarity 0.85)
- Masked Autoencoders for Microscopy are Scalable Learners of Cellular Biology (CVPR 2024 · similarity 0.84)
- Higher-Order Cell Tracking Transformer (NeurIPS 2026 · similarity 0.83)
- FISBe: A Real-World Benchmark Dataset for Instance Segmentation of Long-Range Thin Filamentous Structures (CVPR 2024 · similarity 0.83)
- ViTally Consistent: Scaling Biological Representation Learning for Cell Microscopy (ICML 2025 · similarity 0.82)
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.
