---
license: apache-2.0
---

🎯 CURVAS2025: Bayesian Modeling of Inter-Annotator Variability in UNet-Based PDAC Tumor Segmentation

This repository provides the fine-tuned Bayesian U-Net weights for **Pancreatic Ductal Adenocarcinoma (PDAC) segmentation** in CT scans, developed for the **CURVAS2025 Challenge**.

📦 Files:  
- FT_0.pt – FT_4.pt → Fine-tuned Bayesian U-Net checkpoints (ensemble)  
- M291.pt → Auxiliary pancreas segmentation model (ROI extraction)  

🔗 Code & inference: [github.com/vboussot/CurvasPDACVI](https://github.com/vboussot/CurvasPDACVI)

Used in our challenge paper:  
*"From Deterministic to Probabilistic: Bayesian Modeling of Inter-Annotator Variability in UNet-Based Segmentation"* – CURVAS2025