Sign In

Communications of the ACM

ACM TechNews

The Next Big Step For Ai? Understanding Video


A screenshot from one of the videos in the Moments in Time Dataset.

Massachusetts Institute of Technology (MIT) and IBM researchers have released a vast dataset of video clips painstakingly annotated with details of each action being carried out.

Credit: Moments in Time Dataset

Researchers at the Massachusetts Institute of Technology (MIT) and IBM recently released the Moments in Time Dataset, a massive archive of three-second video clips annotated with minute details of the action being performed.

Recognizing actions effectively will require that machines learn about, for example, a person taking an action and transferring this knowledge to a situation in which an animal is performing the same action, says IBM's Danny Gutfreund.

He notes this technology, known as transfer learning, will be important for the future of artificial intelligence (AI).

The current AI boom was partly sparked by success in teaching computers to recognize the contents of static images by training deep neural networks on large labeled datasets. The next challenge could be teaching machines to understand what a video contains, as well as what is happening in the footage, a breakthrough that could lead to powerful new ways of searching, annotating, and mining videos.

From Technology Review
View Full Article

 

Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA


 

No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account