NVIDIA Research has recently created a new version of GauGAN2, a wildly popular AI painting demo. The deep learning model allows anyone to channel their imagination into photorealistic pictures.
GauGAN user interface. Image credit: NVIDIA
Users can enter a brief phrase to quickly generate the key features and theme of an image. Also, it is possible to generate a segmentation map, a high-level outline showing the location of objects in the scene, and switch to a drawing regime from there.
The demo is one of the first to combine multiple modalities — text, semantic segmentation, sketch, and style — within a single GAN framework. The model behind GauGAN2 was trained on 10 million high-quality landscape images using the NVIDIA Selene supercomputer that is among the world’s ten most powerful supercomputers. The neural network behind GauGAN2 produces a greater variety and higher quality of images than the current state-of-the-art.