The neural network was taught to turn pictures on Instagram into three -dimensional images right in the browser (video) • Newsru.com • RIMA

The neural network was taught to turn pictures on Instagram into three -dimensional images right in the browser (video)

The developer Siril Diane from Google Arts created an extension for the Google Chrome browser, with which you can convert photos on Instagram into three -dimensional and animated images.

According to N+1 , the designed Diane of the extension is based on the explorers from Facebook and the Virginia Polytechnic University of Virginia, which can qualitatively separate objects in the foreground of the image from the background, and then finish the empty sections of the background based on neighboring regions. Diange used the open code of the algorithm and adapted it for Instagram.

Just Pushed the Code of a Chrome Extance that Turns Evel Instagram Into 3D Images using #3DPHOTOOINPINATING . No GPU NEEDED THENKS to @GOOGLECOLAB BUT A bit of patency to set it up ;-)

Demo: @parrstudio 'S Amazing Work
Code: https://t.co/59YJUVRHXE #AIIX #Interation #ML pic.twitter.com/86MmbWDM7V
- Cyril Diagne (@Cyrildiagne) April 19, 2020

The algorithm works on the basis of the neural network. At the same time, the processing of several photos from Instagram at once requires large computational resources. To solve this problem, Diange made this problem that the user part of the development works as an extension for a browser, while image processing occurs in a cloud on Google Colab.

Diange posted the expansion code on GitHub , promising to finalize the instructions for it so that ordinary users who do not have special skills can deal with the program.

Recall that in June last year, a group of researchers from the Imperial College of London and Samsung introduced a neural network capable of "reviving" a photograph of a person, posting on it a record of his own or someone else's voice. In February 2019, NVIDIA published in the public domain code of the Stylegan algorithm, which allows to generate new images using neural networks. Among other things, the program can generate frighteningly realistic faces of non -existent people.