A TensorFlow implementation of CycleGAN for neuroimaging data.
The CycleGAN model is used to convert Post-gadolinium enhanced MRI images into pre-gad MRI images. The main advantage of using CycleGAN is to eliminate the requirement of paired data.
The network architecture can be described as follows:
The generator is formed by using ResNet block defined in models.py. resnet_block is a neural network layer which consists of two convolution layers where a residue of input is added to the output.
The losses used in CycleGAN architecture are:
- Generator Loss
- Discriminator Loss
- Cyclic Loss