Research map / Label-efficient and robust learning / Semi-supervised learning
Two Stream Active Query Suggestion for Active Learning in Connectomics
Zudi Lin, Donglai Wei, Won-Dong Jang, Siyan Zhou, Xupeng Chen, Xueying Wang, Richard Schalek, Daniel Berger, Brian Matejek, Lee Kamentsky, Adi Peleg, Daniel Haehn and 4 more
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 two-stream query suggestion instead of core-set | active learning query suggestion · EM-R50 synapse (Round 2) | ROI proposal accuracy: 0.895 -> 0.926 | Improved |
| use two-stream query suggestion instead of learning-loss | active learning query suggestion · EM-R50 mitochondria (Round 2) | ROI proposal accuracy: 0.771 -> 0.809 | Improved |
| use two-stream query suggestion instead of random sampling | active learning query suggestion · EM-R50 synapse (Round 1) | ROI proposal accuracy: 0.824 -> 0.892 | Improved |
| add hierarchical fusion (N=64, M=4) instead of mask-only clustering (ablation) | two-stream clustering · EM-R50 synapse clustering | label accuracy: 0.819 -> 0.846 | Improved |
| add image stream by late-fusion concatenation to mask-only clustering (ablation) | two-stream clustering · EM-R50 synapse clustering | label accuracy: 0.819 -> 0.738 | Hurt |
| cluster by hierarchy instead of late fusion (ablation) | two-stream clustering · EM-R50 synapse clustering | label accuracy: 0.738 -> 0.846 | Improved |
| add mask stream on top of image-only clustering (ablation) | two-stream clustering · EM-R50 synapse clustering | label accuracy: 0.578 -> 0.819 | Improved |
| use clustering-based annotation instead of random annotation of 4.46% (ablation) | clustering module · EM-R50 synapse clustering | label accuracy: 0.772 -> 0.846 | Improved |
| fine-tune Eu with new labels (ablation) | active clustering · EM-R50 synapse active clustering | label accuracy: 0.846 -> 0.880 | Improved |
| fine-tune Es with new labels (ablation) | active clustering · EM-R50 synapse active clustering | label accuracy: 0.846 -> 0.830 | Hurt |
| also fine-tune Es in addition to Eu (ablation) | active clustering · EM-R50 synapse active clustering | label accuracy: 0.880 -> 0.871 | Hurt |
| more sub-clusters per mask cluster (M=8, N=32 vs M=4, N=64) (ablation) | hierarchical clustering · EM-R50 synapse clustering | label accuracy: 0.846 -> 0.814 | Hurt |
Closest papers by meaning
- Guided Proofreading of Automatic Segmentations for Connectomics (CVPR 2018 · similarity 0.84)
- Boosting Active Learning via Improving Test Performance (AAAI 2022 · similarity 0.83)
- Active Classification with Few Queries under Misspecification (NeurIPS 2024 · similarity 0.82)
- Alleviate and Mining: Rethinking Unsupervised Domain Adaptation for Mitochondria Segmentation from Pseudo-Label Perspective (AAAI 2025 · similarity 0.82)
- ActSort: An active-learning accelerated cell sorting algorithm for large-scale calcium imaging datasets (NeurIPS 2024 · similarity 0.82)
- Diffusion-Driven Two-Stage Active Learning for Low-Budget Semantic Segmentation (NeurIPS 2025 · similarity 0.81)
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.
