Implementation of "Rethinking Conditional GAN training: an approach using geometrically structured latent manifolds." [PDF]
Tensorflow = 1.9 Python = 3.x Scipy = 1.4
The training files should be placed at data_dir/images/training/
The validation files should be placed at data_dir/images/validation/
Each training and validation file should be a concatination of the input and the corresponding output. An example is shown below:
python p2p_geo.py --mode=train --data_dir=data directory --learning_rate=1-e4 --logs_dir=logs directory --batch_size=25
python p2pGeo.py --mode=visualize