/Face-Generator

Uses generative adversarial networks to create images of faces

Primary LanguagePythonMIT LicenseMIT

Face-Generator

This is an example of generative adversarial networks used to generate images of faces. It trained on more than 200,000 images of faces, learning to create new ones. I'm still experimenting with them model architecture, but the results look somewhat decent.

Example of a real face:

Real face

Examples of generated faces:

Generated face1 Generated face2 Generated face3 Generated face4 Generated face5 Generated face6

This model was trained using the CelebA dataset, rescaled to 64x64, and converted to grayscale.