Meet 'SAM,' Meta's New A.I. Segment Anything Model That Can Pinpoint Any Item In Your Photos The new artificial intelligence model has the ability to identify various objects in images and videos.
Meta unveiled its Segment Anything Model "SAM" on Wednesday. The new technology has the ability to identify individual objects in images through what Meta says is the largest segmentation dataset of its kind. The tool can "segment" different items within pictures using a dataset of over 1 billion masks, or various objects, on 11 million licensed and privacy-respecting images, they said in a blog post.
However, they said SAM has been able to identify images it hasn't encountered in its training as well.
Today we're releasing the Segment Anything Model (SAM) — a step toward the first foundation model for image segmentation.— Meta AI (@MetaAI) April 5, 2023
SAM is capable of one-click segmentation of any object from any photo or video + zero-shot transfer to other segmentation tasks https://t.co/qYUoePrWVi pic.twitter.com/zX4Rxb5Yfo
By clicking on objects in images or filling in the text prompts, SAM can identify them for you. For example, in one demonstration SAM drew boxes around several cats in a photo after the word "cat" was entered into the prompt.
Meta has already been utilizing technologies similar to SAM internally for tagging photos, moderating content, and suggesting posts to users. With the release of SAM, Meta hopes to increase access to similar broadscale and adaptable technologies, researchers said in a report.
The SAM model and dataset are available for download under a non-commercial license, and while users can upload their own images, people must agree to use the tool for research purposes only.
It's unclear if Meta checked with the Sams of the world before unveiling the new technology's moniker, however, I, Sam Silverman, did not sign off on this. Here's hoping we Sams don't become the next Siri.