Commit 3d846f4f authored by Guillaume HEUSCH's avatar Guillaume HEUSCH
Browse files

[doc] added DCGAN user guide

parent cac9487c
.. py:currentmodule:: bob.learn.pytorch
DCGAN stands for deeply convolutional GAN, and is described in the following paper [dcgan]_::
Author = {Radford, A. and Metz, L. and Chintala, S.},
Title = {Unsupervised {R}epresentation {L}earning with {D}eep
{C}onvolutional {G}enerative {A}dversarial {N}etworks},
BookTitle = {Intl {C}onf. on {L}earning {R}epresentation},
seq-number = {50},
year = 2016
Also, most of the code (architecture and training) has been borrowed `here <>`_.
Here, the goal is to generate images of frontal faces. You can train a model
using the MULTI-PIE database by using the dedicated script, and a proper configuration
file. The configuration file should be as follow:
.. code-block:: python
### DATA ###
from bob.learn.pytorch.datasets.multipie import MultiPIEDataset
from bob.learn.pytorch.datasets import RollChannels
from bob.learn.pytorch.datasets import ToTensor
from bob.learn.pytorch.datasets import Normalize
import torchvision.transforms as transforms
dataset = MultiPIEDataset(root_dir='path/to/multipie/data',
Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
### NETWORK ###
from bob.learn.pytorch.architectures import DCGAN_generator
from bob.learn.pytorch.architectures import DCGAN_discriminator
from bob.learn.pytorch.architectures import weights_init
ngpu = 1
generator = DCGAN_generator(ngpu)
discriminator = DCGAN_discriminator(ngpu)
.. note::
You should have the ``bob.db.multipie`` package installed on your environment
Once your configuration file is done. You can learn a model to generate
frontal face images by invoking the following script::
$ ./bin/ -vv
If you have access to a GPU, you should run the script using this option::
$ ./bin/ -vv --use-gpu
Note that this script will consider frontal faces only (but all illumination conditions).
Generated samples will be saved after each training epoch. You can specify where to write
images using the ``--output-dir`` option (default to ``./dcgan``).
Here are some examples of saved images.
.. figure:: img/dcgan-multipie-epoch-1.png
:align: center
After 1 epoch
.. figure:: img/dcgan-multipie-epoch-5.png
:align: center
After 5 epoch
.. figure:: img/dcgan-multipie-epoch-20.png
:align: center
After 20 epoch
.. todo:: script that samples from the model
.. [dcgan] *A.Radford, L. Metz, S. Chintala*. **Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks** Intl Conf. on Learning Representation, 2016. `arXiv <>`__
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment