Today, the Nvidia GTC 2019 company unveiled a stunning image creator. Using generic racing networks, software users can sketch images that are almost photorealistic with just a few clicks. The software will instantly turn several lines into a wonderful mountain sundown. This is MS Paint for Age AI.
Called GauGAN, the software is just a demonstration of what is possible with its network platforms. It is designed to draw an image of how a person would draw, the purpose of which is to draw a sketch and become a photorealistic picture in seconds. An early demonstration seems to work as advertised.
GauGAN has three tools: a bucket of paint, a pen and a pencil. At the bottom of the screen there is a series of objects. Select the cloud and draw a line with a pencil, and the software will create dots of photorealistic clouds. But these are not figurative marks. GauGAN gives results unique to the entrance. Draw a circle and fill it with the paint bucket, and the software will make glowing clouds in the summer.
Users can use the input tools to draw the shape of the tree and it will produce a tree. Draw a straight line and it will produce a naked trunk. Draw a bulb at the top and the software will fill it with leaves that produce a full tree.
GauGAN is also multimodal. If two users create the same sketch with the same settings, the built-in random numbers guarantee that the software produces different results.
To have real-time results, GauGAN has to work on a Tensor computing platform. Nvidia demonstrates this software on the RDX Titan graphics platform, allowing it to deliver real-time performance. The demo operator is able to draw a line and the software instantly gives results. However, Bryan Catanzaro, vice president of Applied Deep Learning Research, said with some modifications, GauGAN can work on almost any platform, including processors, although the results may take a few seconds to show up. is not perfect and the team behind the project will improve. There is a light line where two objects are touched. Nvidia calls the results photorealistic but under surveillance does not rise. Currently, neural networks have a problem with the objects they are trained for and what the neural network is trained to do. This project hopes to reduce this gap.
Nvidia turned to 1
Nvidia will release the white paper today. Catanzaro noted that he was previously admitted to the CVPR 2019.
Catanzaro hopes that this software will be available on Nvidia's new AI Playground, but says the company should do some work to do that. He sees tools like this to be used in videogames to create a more compelling environment, but notes that Nvidia does not create software directly for that.
It's easy to complain about the ease with which this software can be used to produce untrue images for criminal purposes. , And Catanzaro agrees that this is an important topic, noting that it is larger than a project and a company. "We are very concerned about this because we want to make the world a better place," he said, adding that this is a matter of trust, not a technological issue, and that we as a society must deal with it as such. 19659002] Even in this limited demonstration, it is clear that the software built around these capabilities will appeal to everyone from the video game designer to the architects to casual gamers. The company does not have any plans to release it, but it may soon launch a public test to allow everyone to use the software.