This notebook is part of the Udacity Deep Learning Nanodegree and it is based from one of their notebooks!
In this project, it is showcased the development of a DCGAN, being composed of a generator and discriminator using convolutional and transpose layers. The model is trained on the CelebA dataset that has been pre-processed to have the faces cropped. More information can be found in the notebook.
The model's goal is to get a generator network to generate new images of faces that look as realistic as possible.
If there are some changes or any ML blasphemies that I committed here and there, feel free to open an issue and point me to the right direction! I'm just learning after all! 😄
The output of this model should produce rather realistic images like the ones that it was trained on. Perhaps more epochs could make a difference to yield more interesting and realistic results but it suffices as a proof of concept.
PyTorch and Google have some amazing references describing GANs and even implementing these.
- The Discriminator by Google
- The Generator by Google
- DCGAN faces walkthrough by PyTorch with great starting hyperparameters.
- Differences between GANs, including different loss functions.
