AI Hub
All papers
Architecture

Segment Anything

Meta·April 5, 2023

Alexander Kirillov, Eric Mintun, Nikhila Ravi

View on arXiv

TL;DR

A promptable segmentation model (SAM) trained on a billion masks, generalizing to new objects zero-shot.

Why it matters

Brought the foundation-model paradigm to computer vision segmentation.

Related terms