February - March 2019

Synthetic Handwritten Digits

As part of a university module, a peer and I trained and tested Generative Adversarial Networks (GAN) to generate synthetic handwritten digit images, utilising the MINST dataset.

We designed various DCGAN's and CGAN's to use as Generator and Discriminator models, then evaluated their performance.

We discussed our findings in a report. All relevant code can be found in the associated GitHub repository.