social network

Netizens missed their photos through an algorithm for “drawing” neural networks from Google

Google has published in the public domain the source code of the algorithm, with which it taught neural networks to “draw” pictures. Users of social networks immediately began to practice using this tool.The Google algorithm, which is designed to automatically find objects in photographs, at a certain setting can, on the contrary, add to the pictures the features of the specified objects.

Runet users were fascinated by the breeding of the eyes in the photo, and also turned the image of shawarma into a work of art.

Western users launched a flash mob and started publishing their “work” created with Google code with the hashtag #deepdream.

There was also a video showing the operation of the algorithm.

A special section dedicated to #deepdream, launched on Reddit. It has already published dozens of images.

Since Google has published only the source code of the algorithm, unqualified users cannot yet conduct their experiments with it. The first attempts to launch online services based on #deepdream appeared in the network, however, such sites immediately stopped working under the influx of volunteers.

About his discovery Google first announced June 17. Employees of the company found that neural networks (complex mathematical models that reproduce the work of the nerve cells of a living organism) created to detect certain objects in photographs can be “turned inside out” and teach them to “draw” complex abstract pictures, analyzing noise and »In it the outlines of various simple objects.

Back to top button