This project was originally started as a final project for an online deep learing course I completed. The course can be found at this url -> https://jovian.ai/learn/deep-learning-with-pytorch-zero-to-gans.

For this project I used the PyTorch framework to build a Generative Adversarial Network (GAN) that takes as input images of all 905 Pokemon in the official Pokedex, and produces new images of fake Pokemon.

A brief overview of the files in this repository:

• PokeScraper.py: The script used to retrieve the images from https://www.pokemon.com/us/pokedex/.
• dataset.py: Contains a custom PyTorch dataset for the Pokemon images.
• device_mgmt.py: Contains a series of helper functions and classes that are used to inspect memory issues or facilitate movement of data between devices.
• models.py: Contains two custom PyTorch models: the generator and the discriminator. The discriminator is a convolutional neural network whose goal is to correctly predict whether images are real or fake, while the generator is a deconvolutional neural network whose goal is to create images that fool the discriminator.
• train.py: A series of helper functions that facilitate the training process for the two networks.
• utils.py: A collection of general helper functions. Utilities for saving predictions, manipulating images, altering file names, etc.
• run.py: The script that brings everything together and starts the training process. In the command line navigate to the directory this repository has been cloned to, then run "python3 run.py". You will be asked for the number of epochs you would like to train for as well as the desired learning rate. The script will run in the command line and save results after each epoch.
• PokeGAN.ipynb: The Jupyter notebook version of this project. The notebook is essentially all the other code in this repository organized into one notebook. This was required for the final submission for my project, and is included for anyone who wants to play around with it.

I don't have any plans to make large adjustments to this project at this point. However, I am always curious about ways I could potentially improve things. If you found this project interesting and care to help me learn, please contact me via email/message if you think I could have done something better. Thanks to anyone who has taken the time to check out my work!