Commit a99cf262 authored by Guillaume HEUSCH's avatar Guillaume HEUSCH
Browse files

[doc] add the need for already preprocessed face images in the GANs doc

parent bc17c6a2
Pipeline #26052 canceled with stage
in 4 minutes and 25 seconds
......@@ -48,10 +48,10 @@ network definition:
from bob.learn.pytorch.datasets import Normalize
import torchvision.transforms as transforms
dataset = MultiPIEDataset(root_dir='/idiap/temp/heusch/data/multipie-cropped-64x64',
dataset = MultiPIEDataset(root_dir='path/to/multipie/data',
frontal_only=False,
transform=transforms.Compose([
RollChannels(), # bob to skimage:
RollChannels(),
ToTensor(),
Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])
......@@ -72,7 +72,9 @@ network definition:
.. note::
You should have the ``bob.db.multipie`` package installed on your environment
You should have the ``bob.db.multipie`` package installed on your environment. Note also
that this model acts on color images of face of size 64x64. It's up to you to preprocess
your data to make them fit these constraints ...
To train the conditional GAN to generate faces with different poses, you should execute::
......
......@@ -53,7 +53,9 @@ file. The configuration file should be as follow:
.. note::
You should have the ``bob.db.multipie`` package installed on your environment
You should have the ``bob.db.multipie`` package installed on your environment. Note also
that this model acts on color images of face of size 64x64. It's up to you to preprocess
your data to make them fit these constraints ...
Once your configuration file is done. You can learn a model to generate
frontal face images by invoking the following script::
......@@ -87,6 +89,4 @@ Here are some examples of saved images.
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 <https://arxiv.org/abs/1511.06434>`__
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