Segment Anything (SAM) is a Meta model from 2023 that outlines any object in an image when you click on it or draw a box around it.
In plain words
Segmentation means marking exactly which pixels belong to an object. SAM is promptable: you give it a point, a box or ask for everything, and it returns a precise outline, called a mask, even for things it never saw in training. Meta built it together with SA-1B, a dataset of over a billion masks on 11 million images, and released both. Later versions, SAM 2 and SAM 3, extended it to video and to text prompts.
A quick example
In a photo of a kitchen, click once on a mug. SAM returns a mask that covers the mug's pixels and nothing else, which you could then cut out or recolour.
Where you'll meet it
- Labelling images quickly to build datasets.
- A starting point for computer vision research.
Related
Sources
- PaperSegment Anything, Kirillov et al. (arXiv) · read 29 Sept 2026
- Docssegment-anything, Meta (GitHub) · read 29 Sept 2026
- Docssam3, Meta (GitHub) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.