The aim of this project is to create a model capable of generating realistic human images that do not exist in reality. The technology behind these kinds of DL model is called a GAN, or Generative Adversarial Network. GANs algorithmic architectures use two networks that work in adversarial manner to create the desired result. The Generator’s goal is to create realistic fake images, while the Discriminator’s goal is to distinguish between real images and fake images in order to guide the Generator to create more realistic images. I did this project as a part of Deep Learning Nanodegree from Udacity.
The dataset used in this project is the Large-scale CelebFaces Attributes (CelebA) Dataset. It consists of colored images of human faces and their annotations. Since we use generative networks, we do not need to the annotations. You can find more information for the full dataset on this link.
- Python 3.7
- Numpy
- Torch
- Torchvision
- Pickle
- Matplotlib
- Jupyter Notebook
- Maybe needs to use GPU
The whole project is located in the jupyter notebook file dlnd_face_generation, you can use the Anaconda environment to open the Jupyter Notebook and install the requirement.