Source: Deep Learning on Medium


To see the full source code and documentation, visit this github repository.


After hours of training on GPU, quickdrawGAN generates these results. They are very cool considering they are generated from noise. The computer (generative network) learns how to draw basketball. You can use the same code for generating other categories doodles available in Quick, draw dataset.

Generative Adversarial Network with time
results after 300 epochs of training