SAM 2 is the Segment Anything Model 2 from Meta AI, a unified foundation model that can segment any object in any image or video in real time. Building on the original Segment Anything Model (SAM), SAM 2 extends promptable segmentation to video with a single model architecture.
Key capabilities
- Promptable segmentation — a click, box, or mask on any image or frame of video.
- Real-time video segmentation — segment and track objects across frames with a single model.
- Memory-aware tracking — the model's memory mechanism keeps objects consistent across the video.
- Zero-shot generalization — works on unseen objects and scenes without additional training.
Getting started
- Try the demo at sam2.metademolab.com.
- Download the model and code from the Segment Anything 2 repository.
- Read the research paper on Meta AI.
SAM 2 is released under the Apache 2.0 license and is available for both research and commercial use.
