Facebook’s image content recognition system has become 10 times better than before

0
14



Usually, search spiders and traditional algorithms do not understand the content of non-textual content, so companies develop artificial intelligence techniques to do this task, and here the capabilities of companies differ, and Facebook is one of the largest parties that invest in the field, and its technologies have become stages better than before.

Facebook has used an artificial intelligence model called AAT since 2016, and now with improved capabilities, it has become ten times better than its beginnings in recognizing the content of images, especially in fine details.

For example, the system can know if a celebrity is in a photo, what he is doing, and even things like hats, instruments, etc. in the image. If you show them a picture of a band, he can tell not just how many people and what they’re wearing, but what instruments they’re playing on and how many.

Facebook can identify famous natural landmarks and animal species in the photos, as well as if people are in the foreground or in the background.

The company used the images posted on public accounts on Instagram, along with the textual description attached to them and the hashtag, to train AI systems and models to improve the accuracy of their content recognition. Even the system has come to distinguish human color and cultural context such as events and holidays.


Source:

Facebook







LEAVE A REPLY

Please enter your comment!
Please enter your name here