Birdy Generative Adversarial Network

Overview

GAN of birdy images. Crawls through google images for each search term (separated by newline) in ./data/search_terms.txt. Set train_data folder in main.py train_data argument to folder containing crawled images, then trains GAN on the data. Outputs generated images in ./data/ folder

Requirements

Python 3.6, Selenium, pattern

Run

main.py [--crawl]  
	[--train_data]  
	[--moden_no]  
	[--batch_size]  
	[--num_epochs]  
	[--lr]  
	[--gradient_acc_steps]  

Results

Left: Training image, Right: Generated image