/Conditional-Face-Generation-with-GANs

Conditional Generative Adversarial Network for generating synthetic faces with user specified attributes

Primary LanguageJupyter NotebookApache License 2.0Apache-2.0

Conditional Face Generation

Generation of faces with specific attributes using Generative Adversarial Networks (GANs) on CelebA dataset

Model

Conditional DCGAN. Labels vector concatenated in the discriminator by means of vectorization. More details can be found in the file cGAN.ipynb

Results

The model achieves a State of the Art FID score of 11.7

  • Unconditional generation

    Alt text

  • Female, Young, Attractive, Brown_Hair, Bangs, Heavy_Makeup

    Alt text

  • Male, Old, Bald, Arched_Eyebrows, Bushy_Eyebrows, Pale_Skin, No_Beard

    Alt text

  • Male, Attractive, Brown_Hair, Straight_Hair, Smiling, Eyeglasses, Mustache

    Alt text

  • Female, Young, Attractive, Blond_Hair, Smiling, Eyeglasses

    Alt text