Progressive Growing of GANs for Improved Quality, Stability, and\n Variation

We describe a new training methodology for generative adversarial networks.\nThe key idea is to grow both the generator and discriminator progressively:\nstarting from a low resolution, we add new layers that model increasingly fine\ndetails as training progresses. This both speeds the training up and greatly\nstabilizes it, allowing us to produce images of unprecedented quality, e.g.,\nCelebA images at 1024^2. We also propose a simple way to increase the variation\nin generated images, and achieve a record inception score of 8.80 in\nunsupervised CIFAR10. Additionally, we describe several implementation details\nthat are important for discouraging unhealthy competition between the generator\nand…