The program can blend hundreds of thousands of objects and their connection to other objects in the real world

Nvidia Research has recently introduced a generative adversarial AI system named GauGAN that allows one to create photorealistic landscapes in seconds. The neural network of GauGAN has been trained with 1 million open source Flickr images and can comprehend the relationship among more than 180 objects like water, snow, flowers, tree, bushes, mountains or hills.

The company revealed that this program can blend hundreds of thousands of objects and their connection to other objects in the real world. GauGAN is apparently just one of the advanced reality-bending AI systems from Nvidia, producer of deepfake tech which can produce realistic images of people who never existed.

With the help of this system, the user can create the shape of a tree with the usage of input tools and it will create a realistic tree. Make a straight line it will create a plain trunk. Also, a tree next to a water source will display a reflection and when there is a change in season, snow would cover the ground and tree will be without leaves.

Reliable sources mentioned that GauGAN creates results that are unique and impressive. With this software, style transfer can also be possible so that an image can show the cooler lights of a city’s skyline or adopt a warm sunset glow.

GauGAN is supposedly a multimodal system, meaning that if two users produce the same sketch with the exact setting, it will ensure that the software creates different results with the random numbers built into the projects.

Sources informed that in order to create real time results, GauGAN has to run on a Tensor computing platform and Nvidia displayed the software on an RDX Titan GPU platform. According to Bryan Catanzaro, VP of Applied Deep Learning Research reported that with some alternations, GauGAN can run efficiently on any platform, including CPUs.

GauGAN builds from the learnings of Pix2Pix system, released last year, which can exhibit virtual worlds, sources added.