Welcome to bob.learn.pytorch's documentation!
This module contains the implementation of different Generative Adversarial Networks (GAN) used for face processing (i.e. generating face images)
If you would like to have an idea on the basic principles of GAN, you should first read the paper by Goodfellow et al.
https://arxiv.org/abs/1406.2661
This package makes heavy use of pytorch, so make sure you have it installed on your environment. It also relies on bob (and in particular for I/O and databases interfaces), so you may want to refer to their respective documentation.
This package is basically organized as follows (some files are omitted for clarity purposes):
bob/
+-- learn/
+-- pytorch/
+-- architectures/
+-- DCGAN.py
+-- ...
+-- datasets/
+-- multipie.py
+-- ...
+-- scripts/
+-- train_dcgan_multipie.py
+-- ...
+-- trainers/
+-- DCGANTrainer.py
-
architectures
contains files defining the different discriminators and generators. -
datasets
contains files implementing the dataset astorch.utils.data.DataSet
, and some utility functions (wrapper around torch.transforms for instance) -
scripts
contains the various scripts to perform training. -
trainers
contains files implementing the different training procedure
In order to build the package, activate your bob environment, and do:
$ buildout
In the user's guide you will find more detailed instructions on how to run the various examples.
References
[dcgan] | A.Radford, L. Metz, S. Chintala. Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks Intl Conf. on Learning Representation, 2016. arXiv |
[cgan] | M. Mirza, S. Osindero. Conditional Generative Adversarial Nets arXiv:1411.1784. |
Licensing Licensing ---------
This work is licensed under the GPLv3.
Indices and tables
- :ref:`genindex`
- :ref:`modindex`
- :ref:`search`