A notebook reflecting the CycleGAN paper in code
- Used tf.data to boost the process in GPU
- dataset used from inbuild tensorflow_dataset's dataset (monet2photo)
- approximately 130 sec's to run 1 epoch when using batches of 4
- image size used is 256x256x3
- checkpoints are created at every 5 epochs, if you don't want to train from scratch then you can use the checkpoint provided here
- pip install tensorflow-dataset
- pip install tensorflow-addons
- Nvidia GPU is prefered to train efficiently