Columbia Engineering researchers unveil a computer vision technique for giving machines a more intuitive sense for what will happen next by leveraging higher-level associations between people, animals, and objects. It is the most accurate method to date for predicting video action events up to several minutes in the future.
After analyzing thousands of hours of movies, sports games, and shows, the system learns to predict hundreds of activities, from handshaking to fist bumping. When it cannot predict the specific action, it finds the higher-level concept that links them. The mathematical framework developed by the researchers enables machines to organize events by how predictable they are in the future.
More information:
https://scienceblog.com/523655/ai-learns-to-predict-human-behavior-from-videos/