braun-steven / food-interpolator

Deep Generative Models Course Project at TU Darmstadt

Food Interpolator

Deep Generative Models Course Project at TU Darmstadt.

Goal

The main goal of this project is to learn a conditional GAN that can interpolate between different types of food. We want to achieve fluid transitions between e.g.:

• Burger <-> Pizza

Data

• Scraped from google with google-images-download
• Partly from PizzaGAN

Results

Video results can be found here.

Progressive Growing (Video)

Pizza to Pizza (Video)

Burger to Burger (Video)

Random Latent Space (Video)

Code Base

The code is based on a PyTorch implementation of Improved Training of Wasserstein GAN and a PyTorch implementation of Progressive Growing of GANs