SimCortex: Pre-trained Models and Dataset Splits

This repository contains the official pre-trained weights and dataset splits for SimCortex V2, a framework for cortical surface reconstruction. Package Contents: seg/seg_best_dice.pt: Pre-trained 3D U-Net weights for 9-class segmentation (Stage 2). deform/deform_best_rmse.pth: Pre-trained network weights for geometric surface deformation (Stage 4). splits/: The train/val/test subject splits (HCP-YA, OASIS) used for model evaluation. For usage instructions, source code, and Docker/Apptainer images, please refer to our main GitHub repository.