Meta has released a new computer vision tool that helps identify objects in images.
The tool includes the new Segment Anything Model (SAM) and Segment Anything 1-Billion mask dataset (SA-1B), Meta said in a Wednesday (April 5) blog post.
According to the post, their name refers to “segmentation”, the process of identifying which image pixels belong to an object.
SAM is available under a permissive open license, while SA-1B is available for research purposes.
According to the post, SAM has learned the general idea of what an object is, so it can identify objects it has never seen before.
“SAM is general enough to cover a wide range of use cases and can be used out of the box with new image ‘domains’ without the need for additional training. ),” Mehta said in a post.
According to the post, the model could help larger artificial intelligence (AI) systems understand the world better, allowing content creators to extract parts of images while creating collages or editing videos. or help scientific researchers study and track animals and objects.
The SA-1B mask dataset contains 1.1 billion masks across different geographic regions, income levels, and demographics to ensure that SAM works equally well for different groups and real-world use cases.
“By sharing our research and datasets, we hope to further accelerate our research on segmentation and image and video understanding more generally,” Meta said in the post.
The field of AI, computer vision, has been used by companies in many ways.
For example, retail tech company Pensa used computer vision and AI to power its shelf scanning solution. This reduces the effort required to manage inventory, improves accuracy, provides more up-to-date information, and increases transparency for customers shopping through e-commerce channels.
Google uses computer vision, AI, and billions of images to create high-fidelity representations of the places featured in the map’s “immersive view.”
Amazon uses computer vision and sensor technology to register products with Just Walk Out cashierless checkout technology.