Python GAN Classifier

Date Completed: July 2020

University Year 3 Term 3

The final part of the 'Advanced Technologies' module presented me with an oppourtunity to build my first Python program to create a neural network capable of learning from a dataset and creating its own addition based off what it had learned. The wider context for this project was to be able to create 2D video game assets that could have randomised splash art to add a unique feel to every object in a game.

The assets I chose to generate with the classifier were footballer profiles, similar to those used in games such as FIFA and Football Manager to represent the players within the UI. A dataset was downloaded, learned, and then new player profiles were replicated.

GAN Classifier Report

To showcase the assets, I quickly built a football stadium scene within Unity and placed the player cards down as a demonstration for how they could be showed within a game setting. The classifier could however be used to generate whatever 2D assets the user wanted though, provided a suitable and substantial dataset was provided, weapons, or potions were two exampels commonly used by others in the cohort.

For more information on the implementation, please read my report linked above

unity1.png