AI to determine items within images
Meta of Facebook stated on Wednesday, that the new AI model is now capable of locating objects inside images. Referred to as “segment anything”, it is said to provide common comprehension of images and videos and accelerates segmentation search.
Segmentation helps people to see, which pixels of an image belong to a specific subject. Meta said it intends to blur barriers used in different applications, like analysis of scientific imagery or photography editing.
In the past it was common to use AI-based infrastructure and big data for high skilled work performance. SAM (Segment Anything Model) and SA-1B have become part of the new project.
They are planned to empower an extensive number of applications and look further into model establishment for computer vision
They say SAM to be utilized for things like underwater photography, as an example, video-editing, imagery extraction from image files.
Unlike images, videos and text, segmentation data required to get this kind of model is not readily available online. So, using Segment Anything we kills two rabbits with one shot: make a segmentation dataset and create a general, prompt segmentation model.
Over 1.1 billion segmentation masks from around 11 million licensed and privacy-preserving images are included in the final dataset.
AI Catalog's chief editor
Share on social networks:
