SAM 2

Segment Anything Model 2 (SAM 2) is a foundation model designed to address promptable visual segmentation in both images and videos. It extends the capabilities of SAM to video by treating images as single-frame videos. The model features a streamlined transformer architecture with streaming memory, enabling real-time video processing. A model-in-the-loop data engine is utilized to refine both the model and its dataset through user interaction, contributing to the development of the SA-V dataset, the largest video segmentation dataset to date. Trained on this extensive dataset, SAM 2 demonstrates strong performance across diverse tasks and visual domains.