Meet SAM… Meta’s latest AI model
New AI model named Sam from Meta just dropped. It stands for Segment anything model, and it is capable of outlining objects in an image in just 50 milliseconds. It was trained on a massive data set of 1 billion image masks. But what makes it unique is that it’s promptable. A user can provide foreground or background points, a bounding box, or just freeform text to determine which objects to segment.
This is huge because it will make extracting objects from images much easier and even for lifting them into 3D in the near future when we all have VR headsets surgically attached to our faces. This tech could quickly identify undesirable objects like ugly faces, then use generative image models to beautify them in real-time or when walking down the streets of San Francisco, homeless people could be easily erased from the landscape, thus making the world a better place.