facebookresearch/sam2

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

19770 stars · 2535 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Official code for Meta's Segment Anything Model 2 (SAM 2), a foundation model for promptable visual segmentation in images and videos. It includes inference code, downloadable model checkpoints, training code, and example notebooks.

Use cases

• segment objects in images with point or box prompts
• track and segment objects across video frames in real time
• download pretrained SAM 2 checkpoints for segmentation tasks
• run video object segmentation with multi-object tracking
• fine-tune or train SAM 2 on custom segmentation data
• extract object masks from images and videos for downstream pipelines

When to choose

• you need state-of-the-art promptable segmentation for images or videos
• you want a foundation segmentation model to build annotation or labeling tools on
• you need real-time video object segmentation with streaming memory
• you want pretrained checkpoints plus training code from the model authors

When to avoid

• you need lightweight segmentation on CPU-only hardware
• you need semantic or instance segmentation with fixed class labels rather than promptable masks
• you need a turnkey application rather than a model library requiring PyTorch and GPUs

Facets

library · maturity active

machine-learning computer-vision image-processing video-processing deep-learning computer-vision machine-learning image-processing artificial-intelligence python cross-platform segment-anything image-segmentation video-segmentation promptable-segmentation foundation-model object-tracking meta-ai inference video gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="facebookresearch/sam2")