CS175 Final Project

Gan Guardians

• Ye Aung
• Arthur Hsu
• Warren Wei Leu
• Tyler Fermanian @TylerFermUCI
• Tariq Brown

Getting Started

Setting Up the Environment

Go to setup/environments and follow the directions to setup your conda environment based on whether your machine is gpu or cpu enabled using Miniconda, a lightweight form of Anaconda.

Downloading Data From AWS

Run the main() function from bps_gan.py to download the images from AWS as .tiff files. Call the separate_particle_type() function to separate different particle types into their own .csv files.

Training the GAN model

• Ensure that the BPSConfig has all the right paths and values(paths to the metadata .csv files are stored, number of epochs, number of batches).
• In main, comment/uncomment lines to define the type of GAN to train (Fe, X-ray, or both).
• Set the width and height of the images, or use the default (128x128).
• Optionally set a path to a checkpoint file (.ckpt) to resume training of an already started (but paused) GAN.
• Uncomment line 392 prepare_data() function before the first run of the file (make sure this line only runs once, or images will be downloaded every file run).
• Run bps_gan.py to train the GAN.