Ever wished that the doodle you drew in your notebook in kindergarten would turn into a beautiful painting? Well, this can become a reality in the coming future.
Nvidia is developing a deep learning model to turn rough doodles into masterpieces with so much ease that drawing doodles will look tougher than being a real, modern day
The app has been named after famous self-portrait artist Paul Gauguin.
GauGAN is aimed to provide a powerful tool for creating virtual worlds to anyone from architect to game developers. With the help of an AI that understands how the real world looks like the developers can make better prototypes and make faster changes in the synthetic world.
GauGAN allows users to sketch down their image and label the different segments as land, sea or air according to the scene they want.
A deep learning program trains a million examples to come up with the end product. In the case of GauGAN, the deep learning model has been trained on millions of images which then fills in the landscape with show-stopping results. Draw water coming down a rock, and it will automatically turn it into a waterfall (which can be seen in the video by Nvidia).
GAN’s consist of a structure of a generator and a discriminator. The generator develops or creates an image and sends it to a discriminator having a database of over a million images instructs the discriminator pixel by pixel to improve the realism of the image. After training on real images, the discriminator knows that water falling down a rock will be a waterfall.
For now, the GauGAN app is useful for nature scenes, but the network underlying this app can be used for buildings, people, vehicles, etc. which are essential for game developing.
That’s all for today, We will keep delivering content every day. Subscribe to GameItToday‘s Newsletter and Notifications (from bottom right) to get great content delivered to you. Till then, Peace. #GameItToday