boyla950 / conditional-wgan-gp

Coursework solutions for a 3rd year Computer Science module on Deep Learning @ Durham University. Uses a conditional WGAN-GP implementation to generate images from the CIFAR-10 and STL-10 datasets.

Deep Learning - Image Generation

This repository contains my solutions to a 3rd year coursework assignment on Deep Learning at Durham University.

Outline

This coursework required the implementation of a Deep Learning model to synthesize images based on the CIFAR-10 and STL-10 datasets. While any method could be used, I opted to implemented a Wasserstein GAN which utilised Gradient Penalty and Auxiliary Classification as described here. The implentation was developed using PyTorch and can be found in the file ac-wgan-gp.py. The results of the model can be seen in the report included in this repository.

Feedback

The final mark received was 84%.

By boyla950.