[ad_1]
Google introduced an experimental service called Move Mirror. This allows you to see your own "reflection" on other people's photos. The service uses videos from a laptop webcam to recognize the position of a user, searches for people at the same place in the database and creates a common GIF animation from that data. .
Developers report that robots can be racial Discrimination
The basis of the service is based on two algorithms, one of which recognizes the pose of a person and creates a simplified model of his skeleton, while the other looks for the most similar posture on many other images. The pose recognition is consistent with the open algorithm of PoseNet, previously developed by Google's experts.
One of its features is that it uses the library TensorFlow.js, which does not require the installation of additional programs on the computer, since all calculations are done via the browser on the # 39; computer of the user. Note that for the normal operation of the service a modern graphics card is required, otherwise the service will process only a few frames per second
Example of body modeled by the PoseNet algorithm
] The second algorithm accepts the model created from the body of the user and looks for the most similar model in the set of photos of the user. ;other people. Developers created 80,000 images from a multitude of videos and use the same PoseNet algorithm tagged the human body model in every shot
Introducing the new Move Mirror service – watch the video
That was fast, the developers used the VP-tree. As a result, on a fairly powerful computer, people performing a search in similar positions take place for tens of milliseconds, so that the display of similar photos in the service takes place in real time. The result of the service can be written as a GIF animation and downloaded to your computer
Also read: ProZorro System received an assistant in the form of artificial intelligence DOZORRO
It should be noted that Google developers do not are not the first to use algorithms to create unusual visual experiences. For example, in 2016, they introduced the Land Lines service. A user takes a line, after which the service finds a similar shaped object on the satellite image, for example a winding shore or a road.
More news related to the news of the world of technology, gadgets and artificial intelligence. Techno
Source link
