Meta has released new computer vision tools that can help you figure out what things are in a picture.
In a blog post on April 5, Meta said that the tools include the new Segment Anything Model (SAM) and the Segment Anything 1-Billion mask dataset (SA-1B).
The post says that their names have to do with “segmentation,” which is the process of figuring out which pixels in an image belong to an object.
The SA-1B can be used for research, while the SAM can be used under a license that lets anyone use it.
According to the post, SAM can figure out what things are because it has a general idea of what they are.
“SAM is general enough to cover a broad set of use cases and can be used out of the box on new image ‘domains’ — whether underwater photos or cell microscopy — without requiring additional training (a capability often referred to as zero-shot transfer),” Meta said in the post.
According to the post, this approach could help larger artificial intelligence (AI) systems understand the world, help content creators extract pieces of images when producing collages or editing movies, and aid scientific researchers study and follow animals or objects.
The SA-1B mask dataset has 1.1 billion masks from different places, income levels, and demographics. This is done so that SAM can work the same way for different groups and in the real world, as the post said.
“By sharing our research and dataset, we hope to further accelerate research into segmentation and more general image and video understanding,” Meta said in the post.
Companies have used computer vision, which is a type of artificial intelligence, in a number of ways.
Retail technology Pensa uses computer vision and AI to power a shelf scanning system that reduces labor, increases accuracy, provides more up-to-date information, and improves transparency for eCommerce customers.
Google’s “immersive view” on maps uses computer vision, AI, and billions of images to show places in a way that is very accurate.
Amazon’s Just Walk Out technology doesn’t need a cashier because it uses computer vision and sensors to register items.
