Timo Karras

ProGAN, or Progressively Growing GAN, is a generative adversarial network that utilises a progressively growing training approach. The idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses.

Timo Karras is a renowned researcher in the field of generative modeling and computer graphics. He is currently a research scientist at NVIDIA focused on generative AI. Karras is best known for pioneering work in high-quality image synthesis using neural networks. He developed the ProGAN architecture in 2017 - one of the first models to generate realistic facial images. Then he developed StyleGAN in 2019 which improved synthesis quality and control using an embedded style space. StyleGAN2 in 2020 added architectural changes like adaptive discriminator augmentation and advanced training techniques.

His recent work includes DALL-E, a text-to-image generation model, and DALL-E 2 which can create photorealistic images from text captions. Karras introduced major improvements to diffusion models like DPM++ that achieves state-of-the-art image generation quality. His papers are widely influential in the generative modeling field and have enabled breakthroughs in synthesizing realistic human faces.

Karras is known for applying rigorous quantitative evaluation of image synthesis quality using metrics like FID. He has contributed enormously to NVIDIA's leadership in developing capable generative AI systems.

Timo Karras is a pioneering researcher who has developed many of the most advanced and high-quality generative neural network models for images and beyond. His work defines the state-of-the-art.