SR-GAN-Flow

CNN-GAN for super-resolution + denoising of MRI blood flow fields

Training

Edit and run GAN-trainer.py or trainer.py to train GAN variants or 4DFlowNet respectively. Architecure settings can be changed in the chosen model's TrainerController.py and SR4DFlowGAN.py

All relative paths assume src/ is the current working directory.

Prediction

Edit and run GAN-predictor.py or predictor.py to predict using GAN variants or 4DFlowNet respectively. Architecure settings must match those used during training for the model weights to load successfully.

All relative paths assume src/ is the current working directory.