Research map / Detection and segmentation / Weakly supervised detection and segmentation
COIN: Confidence Score-Guided Distillation for Annotation-Free Cell Segmentation
Sanghyun Jo, Seo Jin Lee, Seungwoo Lee, Seohyung Hong, Hyungseok Seo, Kyungsu Kim
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 unsupervised semantic segmentation (USS) propagation (ablation) | SSA baseline · MoNuSeg train | FN rate: 0.536 -> 0.020 | Improved |
| add USS propagation without OT (ablation) | SSA baseline · MoNuSeg train | FP rate: 0.159 -> 0.552 | Hurt |
| add optimal transport after USS propagation (ablation) | USS propagation · MoNuSeg | FP rate: 0.552 -> 0.303 | Improved |
| replace GMM clustering with optimal transport (ablation) | USS + clustering · MoNuSeg test | IoU: 0.482 -> 0.539 | Improved |
| add CRF after USS (ablation) | USS propagation · MoNuSeg | IoU: 0.439 -> 0.443 | No clear effect |
| add watershed post-processing (ablation) | USS + OT + CRF · MoNuSeg train | AJI: 0.001 -> 0.380 | Improved |
| add instance-level scored self-distillation (ablation) | SSA baseline · MoNuSeg test | AJI: 0.259 -> 0.386 | Improved |
| add pixel-level cell propagation on top of scored distillation (ablation) | COIN · MoNuSeg test | AJI: 0.386 -> 0.580 | Improved |
| add instance propagation without scored distillation (ablation) | SSA baseline · MoNuSeg test | AJI: 0.259 -> 0.260 | No clear effect |
| use SAM masks as pseudo-GT without confidence scoring (ablation) | SSA baseline · MoNuSeg test | AJI: 0.259 -> 0.000 | Hurt |
| add instance-level confidence scoring to SAM pseudo labels (ablation) | SSA + SAM · MoNuSeg test | AJI: 0.000 -> 0.580 | Improved |
| add edge decoder (ablation) | COIN (SSA) · MoNuSeg test adjacent cells | AJI: 0.493 -> 0.528 | Improved |
Closest papers by meaning
- Supervise Less, See More: Training-free Nuclear Instance Segmentation with Prototype-Guided Prompting (ICML 2026 · similarity 0.87)
- Disco: Densely-overlapping Cell Instance Segmentation via Adjacency-aware Collaborative Coloring (ICLR 2026 · similarity 0.87)
- The Four Color Theorem for Cell Instance Segmentation (ICML 2025 · similarity 0.86)
- Weakly Supervised Gland Segmentation with Class Semantic Consistency and Purified Labels Filtration (AAAI 2025 · similarity 0.84)
- The Devil is in the Object Boundary: Towards Annotation-free Instance Segmentation using Foundation Models (ICLR 2024 · similarity 0.83)
- Mudslide: A Universal Nuclear Instance Segmentation Method (CVPR 2024 · 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.
