Meta's SAM Revolutionizes Image Segmentation with AI
Meta put the AI community on notice Wednesday officially announcing its artificial intelligence model, SAM (Segment Anything Model).
SAM can precisely identify objects in images and videos, even if it hasn't encountered them during training. Alongside the model, Meta released an extensive dataset of image annotations, considered the largest ever of its kind.
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
SAM enables users to select objects by either clicking on them or writing text prompts, streamlining the image segmentation process. For example, typing "cat" prompts the AI to outline each cat in a photo. This development comes as tech giants compete in the AI space, following the success of Microsoft-backed OpenAI's ChatGPT chatbot. Meta has been exploring generative AI, akin to ChatGPT, which creates new content rather than just identifying or categorizing data.
Meta CEO Mark Zuckerberg has committed to incorporating generative AI "creative aids" into Meta's applications this year. The technology underpinning SAM has been used internally for tagging photos, moderating prohibited content, and recommending posts on Facebook and Instagram. The release of SAM is aimed at broadening access to such advanced technology.
Meta's SAM model and dataset will be available for download under a non-commercial license. Users uploading their own images to the accompanying prototype must agree to use it exclusively for research purposes, promoting an environment of innovation and exploration within the AI community.