Meta publishes AI model SAM, which can accurately identify objects in the picture.
CTOnews.com April 6 news, Meta announced today the launch of the Segment Anything tool, which can accurately identify objects in images. The project includes models, datasets, and allows others to use it under a looser Apache 2.0 license.
The model is called Segment Anything Model (SAM), and its dataset is called Segment Anything 1-Billion mask dataset (SA-1B), which Meta officials say is the largest split dataset (Segmentation Dataset) ever.
CTOnews.com learned from the official Meta press release that the SAM model can be used in a variety of fields to find and segment any object in an image.
In the technology of this model, developers can develop artificial intelligence with more powerful function and wider range of influence. In virtual reality (VR) and augmented reality (AR) scenes, the relevant information of objects can be displayed interactively according to the eyes of users.
Related papers: https://ai.facebook.com/research/publications/segment-anything/
Press release: https://ai.facebook.com/blog/segment-anything-foundation-model-image-segmentation/
Demo page: https://segment-anything.com/