Training generative adversarial networks (GANs) on high quality (HQ) images
involves important computing resources. This requirement represents a
bottleneck for the development of applications of GANs. We propose a transfer
learning technique for GANs that significantly reduces trainin