MIT-IBM developed a faster way to train video recognition AI

    Date09 Oct 2019
    1432
    Posted ByBrittany Day
    LS Hmepg 337x500 24

    A team from MIT-IBM has developed a faster way to train video recognition AI, which could make it easier to run machine learning on mobile devices.

    Machine learning has given computers the ability to do things like identify faces and read medical scans. But when it's tasked with interpreting videos and real-world events, the models that make machine learning possible become large and cumbersome. A team from the MIT-IBM Watson Lab believe they have a solution. They've come up with a method that reduces the size of video-recognition models, speeds up training and could improve performance on mobile devices.

    The trick is in shifting how video recognition models view time. Current models encode the passage of time in a sequence of images, which creates bigger, computationally-intensive models. The MIT-IBM researchers designed a temporal shift module, which gives the model a sense of time passing without explicitly representing it. In tests, the method was able to train the deep-learning, video recognition AI three times faster than existing methods.

    You are not authorised to post comments.

    Comments powered by CComment

    LinuxSecurity Poll

    What do you think of the articles on LinuxSecurity?

    No answer selected. Please try again.
    Please select either existing option or enter your own, however not both.
    Please select minimum 0 answer(s) and maximum 3 answer(s).
    /main-polls/24-what-do-you-think-of-the-quality-of-the-articles-on-linuxsecurity?task=poll.vote&format=json
    24
    radio
    [{"id":"87","title":"Excellent, don't change a thing!","votes":"67","type":"x","order":"1","pct":57.26,"resources":[]},{"id":"88","title":"Should be more technical","votes":"16","type":"x","order":"2","pct":13.68,"resources":[]},{"id":"89","title":"Should include more HOWTOs","votes":"34","type":"x","order":"3","pct":29.06,"resources":[]}]["#ff5b00","#4ac0f2","#b80028","#eef66c","#60bb22","#b96a9a","#62c2cc"]["rgba(255,91,0,0.7)","rgba(74,192,242,0.7)","rgba(184,0,40,0.7)","rgba(238,246,108,0.7)","rgba(96,187,34,0.7)","rgba(185,106,154,0.7)","rgba(98,194,204,0.7)"]350
    bottom200

    We use cookies to provide and improve our services. By using our site, you consent to our Cookie Policy.