# This script will train image segmentation models in different dataset configurations as in the paper

import Image2Image

opt = Image2Image.get_opt()

for dataset_name in ["cityscapes", "edges2shoes"]: # Iterate over datasets, more datasets could be added here like maps

opt.dataset = dataset_name

for num_joint_samples in [100, 1000, 10000]: # Try for different amount of paired samples

# Apply settings

print(str(num_joint_samples) + " joint samples")

opt.num_joint_samples = num_joint_samples

print("Training GAN")

opt.experiment_name = str(num_joint_samples) + "_joint_GAN"

opt.factorGAN = 0

Image2Image.train(opt)

print("Training factorGAN")

opt.experiment_name = str(num_joint_samples) + "_joint_factorGAN"

opt.factorGAN = 1

Image2Image.train(opt)