A U-Net based end-to-end MRI motion correction reimplementation in pytorch. With/Without GAN version is available.
The basic idea and the motion simulation is based on the MRM article Conditional generative adversarial network for 3D rigid-body motion correction in MRI and the Github code MoCo_cGAN. And there is a lot of code reused from ResNet50-Unet.
The scheduler position is updated for higher version (higher than 1.1.0) of pytorch. Please check out the official warning.
This project can also be used as a template for any U-Net based task and (patch) GAN-based task.