This project was developed during the Machine Learning for Medial Imaging practical during Summer Semester 2019 which was offered by the Chair for Computer Aided Medical Procedures & Augmented Reality (CAMP) from the Technical University Munich (TUM).
The goal of this project was to implement a deep learning-based automatic registration of the human pelvis from a 2D X-Ray scans to estimate the pose of a patient and then automatically register the 3D CT scan. Our implementations were based on the Stacked Hourglass Network and the Convolutional Pose Machines. We applied several changes to improve the performance of the proposed networks.
This project is based on Pytorch-Template by Victor Huang