GP-Optimized Graph Cut for Training-Free Nuclei Segmentation

This archive contains the full implementation, notebooks, and experimental outputs associated with the manuscript: "Genetic Programming Evolving Robust Unary Costs for Efficient Graph Cut Segmentation" The materials include:- The complete GP-optimized graph cut pipeline for training-free nuclei segmentation- A lightweight U-Net baseline notebook used for contextual comparison- Per-image quantitative results (Dice, IoU, Precision, Recall, F1) in CSV format- Automatic foreground and background seed locations in JSON format- All configuration details required to reproduce the reported experiments The framework performs per-image optimization and does not require GPU training or large annotated datasets. It is validated on BBBC038 fluorescence microscopy images and MoNuSeg histopathology images. These resources enable full reproduction of the results reported in the manuscript.