Robust Conditional GAN

Chainer implementation of the ICLR paper "Robust Conditional Generative Adversarial Networks"

https://openreview.net/forum?id=Byg0DsCqYQ [1]

https://link.springer.com/article/10.1007/s11263-020-01348-5 [4]

Robust Conditional GANs aim at leveraging structure in the target space of the generator by augmenting it with a new, unsupervised pathway to learn the target structure.

Testing/demo mode

We provide a jupyter notebook that illustrates how to call the code for testing.

If you want to use the pretrained models, please follow the instructions in the notebook. The demo testing images are provided in the demo/ folder.

Train the network

To train the network, e.g. with rocgan, you can execute the following command:

python jobs/rocgan/train_mn.py '--config jobs/rocgan/iclr_5layer_rocgan_super.yml'

The yml file describes the modules/dataset to train on. The hyper-parameters are included in the yml, no need to hardcode them in the files.

Dataset preparation

The code is tuned for super-resolution (or any task that requires a pair of input/output images).

In each iteration, a single file is loaded from disk; this file is the vertical concatenation of the input and the output images. Please see the demo/ folder for some samples; the idea is to vertically concatenate the images with the input (e.g. corrupted) image on top and the output image on the bottom. All images are resized to 64x64 for training/testing.

Browsing the folders

The folder structure is the following:

• gen_models: The folder for the generator models.
• dis_models: The folder for the discriminator models; do not modify the optional arguments, unless you know what you are doing.
• updater: The folder that contains the core code for the updater, i.e. the chunk of code that runs in every iteration to update the modules.
• jobs: It contains a) the yml with the hyper-parameter setting, b) the main command to load the models and run the training (train_mn*.py).
• source: It contains auxiliary code, you should probably not modify any of that code.
• evaluations: It contains the code for validation (either during training or offline).

Misc

The results and the pretrained files correspond to the IJCV version [4].

Tested on a Linux machine with:

• chainer=4.0.0, chainercv=0.9.0,
• chainer=5.2.0, chainercv=0.12.0.

The code is highly influenced by [2].

Apart from Chainer, the code depends on Pyaml [3].

Citing

If you use this code, please cite [1] or [4]:

BibTeX:

@inproceedings{
chrysos2018rocgan,
title={Robust Conditional Generative Adversarial Networks},
author={Grigorios G. Chrysos and Jean Kossaifi and Stefanos Zafeiriou},
booktitle={International Conference on Learning Representations (ICLR)},
year={2019},
url={https://openreview.net/forum?id=Byg0DsCqYQ},
}

or:

@article{
chrysos2020rocgan,
title={RoCGAN: Robust Conditional GAN},
author={Grigorios G. Chrysos and Jean Kossaifi and Stefanos Zafeiriou},
journal={International Journal of Computer Vision},
pages={1--19},
year={2020},
publisher={Springer}
}

References