Research map / Detection and segmentation / CNN features
Guided Proofreading of Automatic Segmentations for Connectomics
Daniel Haehn, Verena Kaynig, James Tompkin, Jeff W. Lichtman, Hanspeter Pfister
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 Guided Proofreading's boundary-context CNN instead of Focused Proofreading's classifier | automatic split-error classification · Classifier test set | test accuracy: 0.68 -> 0.94 | Improved |
| Rank candidate corrections with Guided Proofreading rather than Focused Proofreading | novice proofreading · AC4 subvolume | median VI: 0.469 -> 0.424 | Improved |
| Rank candidate corrections with Guided Proofreading rather than Focused Proofreading | novice proofreading · AC4 subvolume | VI reduction per minute: 0.00023 -> 0.00173; $7.5\times$ | Improved |
| Rank candidate corrections with Guided Proofreading rather than Focused Proofreading | expert proofreading · AC4 subvolume | median VI: 0.439 -> 0.396 | Improved |
| Use Guided Proofreading rankings rather than Focused Proofreading rankings | oracle-selected corrections · AC4 subvolume | median VI: 0.353 -> 0.342 | Improved |
| Use Guided Proofreading rankings rather than Focused Proofreading rankings | automatic selection at threshold $p_t=0.95$ · AC4 subvolume | median VI: 1.9 -> 0.398 | Improved |
| Use Guided Proofreading rankings rather than Focused Proofreading rankings | automatic selection at threshold $p_t=0.95$ · L. Cylinder | median VI: 2.75 -> 0.352 | Improved |
| Use Guided Proofreading instead of Focused Proofreading to rank corrections | selection oracle · L. Cylinder | median VI: 0.298 -> 0.2996 | No clear effect |
| Add candidate merge-error detection and corrections | Guided Proofreading · AC4 subvolume | mean VI: 0.512 -> 0.509 | Improved |
| Add candidate merge-error detection and corrections | Guided Proofreading · L. Cylinder, 50 sections | VI effect | No clear effect |
Closest papers by meaning
- ConnectomeBench: Can LLMs proofread the connectome? (NeurIPS 2025 · similarity 0.89)
- An Error Detection and Correction Framework for Connectomics (NeurIPS 2017 · similarity 0.89)
- Biologically-Constrained Graphs for Global Connectomics Reconstruction (CVPR 2019 · similarity 0.88)
- Large Scale Image Segmentation with Structured Loss Based Deep Learning for Connectome Reconstruction (TPAMI 2018 · similarity 0.85)
- Pancakes: Consistent Multi-Protocol Image Segmentation Across Biomedical Domains (NeurIPS 2025 · similarity 0.83)
- Rethinking Decoder Design: Improving Biomarker Segmentation Using Depth-to-Space Restoration and Residual Linear Attention (CVPR 2025 · 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.
