GauGAN – Semantic Image Synthesis with GANs

Datum konání: 29.03.2019
Přednášející: Zuzana Bílková
Odpovědná osoba: Kotera

Nvidia recently introduced a new artificial intelligence application GauGAN that turns segmentation maps into photorealistic images. We will present its deep learning model leveraging GANs (generative adversial networks). GAN is a class of machine learning systems with two neural networks, generator and discriminator, contesting with each other. The generator presents images to the discriminator which is trained to distinguish them from real images.