Communications of the ACM
AI Learns to Predict Human Behavior From Videos
Columbia Engineering researchers have developed a computer vision technique to give machines a more intuitive sense for what will happen next by leveraging higher-level associations between people, animals, and objects.
Credit: Didac Suris/Columbia Engineering
A computer vision algorithm developed by Columbia Engineering researchers can intuitively predict human interactions and body language in video, using a mathematical framework that enables machines to organize events by their predictability.
After it has analyzed thousands of hours of movies, sports events, and TV shows, the system learns to anticipate hundreds of actions; when predicting a specific action is impossible, it finds the higher-level concept connecting them.
The researchers say the algorithm is the most accurate technique to date for forecasting video action events several minutes in advance.
Columbia Engineering's Didac Suris said, "When a person cannot foresee exactly what will happen, they play it safe and predict at a higher level of abstraction. Our algorithm is the first to learn this capability to reason abstractly about future events."
From Columbia Engineering
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found