Srujan35007 / Deep-Generative-Modelling

Deep generative models using Generative Adversarial Networks(GANs).

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Deep-Generative-Modelling

Generative models using Deep convolutional Networks.

Data collection and pre-processing

  • CelebA dataset by Jessica Li and team.
  • Cropped to square and resized to (160, 160, 3). (Gray scale images can also be used if computation is expensive).
  • Stored in a zip file without compromizing the file names.
  • Refer preprocess_images.ipynb for more details.

Here is a collage of preprocessed images

collage

The model architecture

  • An up-sampling convolutional network for the Generator.
  • A down-sampling convolutional network for the Discriminator.

Generator architecture

generator

Discriminator architecture

discriminator

About

Deep generative models using Generative Adversarial Networks(GANs).

License:MIT License


Languages

Language:Jupyter Notebook 100.0%