Meta Unveils Game-Changing Object Detection AI Model “SAM” and Unmatched Dataset
Facebook’s parent company, Meta, has announced its latest foray into the world of artificial intelligence. Meta’s new AI model, named “Segment Anything Model,” can identify and isolate individual objects within an image or video with remarkable accuracy.
According to Meta, the “Segment Anything Model” is more than just a tool for content creation and photo editing; it can potentially transform how we see and interact with the world around us. The tool’s ability to identify individual objects within an image opens up a host of possibilities for industries ranging from AR (Augmented Reality) to scientific research.
In addition to unveiling its new AI model, Meta has also made available a groundbreaking dataset of image annotations. This dataset, which Meta claims is the most comprehensive of its kind, includes annotations for a vast array of objects and scenarios, from everyday household items to complex scientific imagery.
Meta’s SAM AI Model Can Recognize Objects it Hasn’t Been Trained on
What sets SAM apart is its ability to identify objects it has not been trained on, giving it an edge over other AI models in the market. Users can choose items inside a video or image using SAM’s simple and user-friendly interface by tapping them or responding to text prompts.
During a recent demonstration, the new AI tool was put to the test with a photo containing several cats. To see how accurately the tool could identify objects, the user simply typed in the word “cat” as a prompt. Within seconds, the tool detected and drew boxes around each of the cats in the photo, showcasing its remarkable ability to easily identify objects.
Read Also: Say Hello To Chat GPT 4 – A Smarter AI Bot
How SAM Works?
SAM is a foundation model that can perform zero-shot and few-shot learning for new datasets and tasks using prompting techniques. The model can return a valid segmentation mask for any prompt, including foreground/background points, rough boxes, masks, and freeform text. The model needs to run in real-time on a CPU in a web browser to allow annotators to use SAM interactively. SAM uses an image encoder, a lightweight encoder, and a lightweight decoder to predict segmentation masks in just 50 milliseconds given any prompt in a web browser.
What Is Meta Saying About SAM
In order to tag photos, moderate material that is prohibited, and suggest posts to people on Instagram and Facebook, the company has been employing technology similar to SAM. The release of SAM will expand access to such technology to a broader audience. The new AI model and dataset are accessible for download from the company’s website under a non-commercial license. Applicants of the demo must consent to only using the tool for research reasons when they upload their images.
Moreover, the SAM model has potential applications in scientific research. It could help researchers identify and monitor specific objects or animals in video recordings of natural phenomena occurring in space or on Earth.
For more of such latest tech news, listicles, troubleshooting guides, and tips & tricks related to Windows, Android, iOS, and macOS, follow us on Facebook, Instagram, Twitter, YouTube, and Pinterest.