Home

Awesome

Estimating the Success of Unsupervised Image to Image Translation

Pytorch implementation of "Estimating the Success of Unsupervised Image to Image Translation" (arxiv).

Prerequisites

Download dataset

Download dataset [edges2shoes, edges2handbags, cityscapes, maps, facades]: bash datasets/download_pix2pix.sh $DATASET_NAME.

General GAN Bound (Alg.1 and Alg.2)

DiscoGAN: python ./discogan_arch/general_gan_bound_discogan.py --task_name=$DATASET_NAME

DistanceGAN: python ./discogan_arch/general_gan_bound_distancegan.py --task_name=$DATASET_NAME

Per Sample Bound (Alg.3)

Train G_1 model:

DiscoGAN: python ./discogan_arch/disco_gan_model.py --task_name=$DATASET_NAME --num_layers=3

DistanceGAN: python ./discogan_arch/general_gan_bound_distancegan.py --task_name=$DATASET_NAME

Then Train G_2:

DiscoGAN: python ./discogan_arch/gan_bound_per_sample_discogan.py --task_name=$DATASET_NAME --pretrained_generator_A_path='./models/model_gen_A-10' --pretrained_generator_B_path='./models/model_gen_B-10' --pretrained_discriminator_A_path='./models/model_dis_A-10' --pretrained_discriminator_B_path='./models/model_dis_B-10' --one_sample_index=$SAMPLE_NUMBER

DistanceGAN: python ./discogan_arch/gan_bound_per_sample_distancegan.py --task_name=$DATASET_NAME --pretrained_generator_A_path='./models/model_gen_A-10' --pretrained_generator_B_path='./models/model_gen_B-10' --pretrained_discriminator_A_path='./models/model_dis_A-10' --pretrained_discriminator_B_path='./models/model_dis_B-10' --one_sample_index=$SAMPLE_NUMBER

Options

Additional options can be found in ./discogan_arch/discogan_based_options/options.py

For specific configuration see DistanceGAN and DiscoGAN

Reference

If you found this code useful, please cite the following paper:

@inproceedings{Benaim2018EstimatingTS,
  title={Estimating the Success of Unsupervised Image to Image Translation},
  author={Sagie Benaim and Tomer Galanti and Lior Wolf},
  booktitle={ECCV},
  year={2018}
}

Acknowledgements

This project has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant ERC CoG 725974).