One motive I’ve been underwhelmed by AI is that firms constantly body it as an answer to each drawback below the solar. That’s why Meta’s new Segment Anything Model (SAM 2) is so intriguing to me. SAM 2 doesn’t reply questions, write code, generate AI photographs, or compose music. As a substitute, as its identify suggests, the brand new AI mannequin merely segments objects in video and pictures — however it does its one job actually, rather well.
Meta describes SAM 2 as “the primary unified mannequin for real-time, promptable object segmentation in photographs and movies.” You should utilize it to pick out a number of objects that may be tracked in actual time, even when they’re shifting round erratically in a video.
Listed here are just some thrilling examples of the brand new Phase Something Mannequin in motion:
In response to Meta, SAM 2’s enhancements over the primary mannequin embrace higher segmentation in photographs, higher monitoring in movies, and thrice much less interplay time than current interactive video segmentation strategies.
If you wish to strive it out for your self to see what the AI mannequin is able to, Meta has a demo on its site that allows you to observe a number of objects in a brief video after which make numerous edits, together with altering the background and including results to the chosen objects.
If you wish to learn extra about SAM 2, take a look at Meta’s latest blog post.
