Computer facial recognition systems already work quite well and even know how to discern emotions. But people transmit information not only through their faces, but also with the help of various gestures and postures. So scientists from the Institute of Robotics at Carnegie Mellon University decided to create a program that would easily “read” the language of human gestures and interpret the information received.
In their development, scientists have involved 500 high-precision video cameras installed inside a huge dome with a height of 2 floors. Five hundred cameras (provided by the Panoptic Studio researchers) produce a huge amount of information, and even when shooting a single frame, considerable computational resources are required to analyze all visual information. At the same time, the system should conduct a simultaneous assessment of the facial expression, position of the head, torso, legs and all fingers. As the leader of the project Yasser Sheikh,
“A person expresses his state and emotions with the help of a facial expression, movements and posture of the body no worse than with his voice. But computers until recently remained "blind" in relation to this language of our body. We managed to “teach” artificial intelligence to what we can do almost from birth. Now we are planning to improve the software so that only one camera is enough for analysis of the system. At the same time, we want to improve the system so that it can decipher the body language of a whole group of people interacting with each other. ”
Data on the new method of deciphering the human body language, as well as its source code, is now freely available. Scientists plan to present more detailed information about their development at the Computer Vision and Pattern Recognition Conference (CVPR) 2017, which will be held from July 21 to 26. Well, you can familiarize yourself with the new product right now with the help of the video below.
The article is based on materials .
Comments
Post a Comment