DCGAN-for-generating-human-faces-

1. Project overview :

In this projects I built a Deep convolutional generative adversarial network (DCGAN) to generate new fake images of human faces .
The model was trained on (CelebA)](http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html)

2. Getting started :

In order to reproduce the same results(that can be found in the notebook itself) , make sure to follow the following steps :

1. Clone the repository :

git clone https://github.com/ZSoumia/DCGAN-for-generating-human-faces-.git

2. Download the dataset :

(CelebA)](http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html)

3. Make sure to use the same tools :

• Pytorch.
• Python 3.6.

Advices :

GANs are a bit too sensitive so trainning them to reach a low loss is a bit challenging, these are some tips that I discovered from training this neural network :

• The BCEWithLogits performed better than the mean of squared errors
• In the optimizer's parameters generally setting beta1 to have a value between 0.2 and 0.35 seemed to improve my results
• Setting the learning rate of the discriminator to be 4 times greater than the learning rate of the generator helped speeding up the training process
• Setting a mini batch to a small value like 16 or 32 also enhanced the results .

License :

This project is licensed under the GNU LICENSE file for details.