Research map / Detection and segmentation / Open-set and few-shot detection
FreeCOS: Self-Supervised Learning from Fractals and Unlabeled Images for Curvilinear Object Segmentation
Tianyi Shi, Xiaohuan Ding, Liang Zhang, Xin Yang
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 PSAL feature alignment on top of FFS (ablation) | FreeCOS · XCAD | Dice: 0.615 -> 0.633 | Improved |
| add image-level intensity-order alignment (IA) (ablation) | FFS+PSAL · XCAD | Dice: 0.633 -> 0.647 | Improved |
| add CMCL (full GIA) (ablation) | FFS+PSAL+IA · XCAD | Dice: 0.647 -> 0.661 | Improved |
| add FDA-based background synthesis (ablation) | FFS · XCAD | Dice: 0.459 -> 0.615 | Improved |
| add Gaussian blur to synthetic images (ablation) | FFS · XCAD | Dice: 0.547 -> 0.615 | Improved |
| vary fractal width (ablation) | FFS · XCAD | Dice: 0.354 -> 0.615 | Improved |
| vary fractal length (ablation) | FFS · XCAD | Dice: 0.358 -> 0.615 | Improved |
| vary branching angle (ablation) | FFS · XCAD | Dice: 0.415 -> 0.615 | Improved |
| vary fractal intensity (ablation) | FFS · XCAD | Dice: 0.569 -> 0.615 | Improved |
| increase synthetic fractal images from 15 to 150 (ablation) | FreeCOS · XCAD | Dice: 0.629 -> 0.661 | Improved |
| increase unlabeled target images from 162 to 1620 (ablation) | FreeCOS · XCAD | Dice: 0.620 -> 0.661 | Improved |
| FFS synthetic training alone versus adversarial self-supervised SSVS | FFS · XCAD | Dice: 0.557 -> 0.615 | Improved |
Later papers that compared with this method
1 later accepted paper reports FreeCOS as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- CaPro: Curvilinear-aware Prompt Learning with Single Unlabeled Image for Cost-effective Curvilinear Structure Segmentation (AAAI 2026 · better on 6 of 6 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- CaPro: Curvilinear-aware Prompt Learning with Single Unlabeled Image for Cost-effective Curvilinear Structure Segmentation (AAAI 2026 · similarity 0.87)
- Enriching Information and Preserving Semantic Consistency in Expanding Curvilinear Object Segmentation Datasets (ECCV 2024 · similarity 0.86)
- DeepBranchTracer: A Generally-Applicable Approach to Curvilinear Structure Reconstruction Using Multi-Feature Learning (AAAI 2024 · similarity 0.85)
- Fractal Calibration for Long-tailed Object Detection (CVPR 2025 · similarity 0.83)
- EFDTR: Learnable Elliptical Fourier Descriptor Transformer for Instance Segmentation (ICML 2025 · similarity 0.83)
- unMORE: Unsupervised Multi-Object Segmentation via Center-Boundary Reasoning (ICML 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.
