Computational Model Studio · Model release
SAM 2.1 Hiera Large
Segments prompted objects in images and follows their masks through video frames.
Developer: Meta · Last verified:
What it is designed to do
Produce and propagate object masks from image or video frames using point or box prompts.
- Primary contribution
- perception
- Input
- image, video, spatial
- Output
- image, video (segmentation masks, tracked video masks)
Documented uses: Prompted object segmentation in still images.; Propagating object masks across video frames.
Access and deployment
- SAM 2.1 Hiera Large model files — available. Public publisher repository; review applicable terms before use. Artifact documentation Local hardware requirements have not been verified. The full artifact-specific rights document has not been separately reviewed; consult the publisher repository conditions.
Technical profile and review notes
- Construction
- learned
- Representation
- neural
- Normalized architecture tags
- No normalized architecture tag assigned
- Lifecycle
- Unknown from reviewed sources
Inclusion rationale: Adds direct visual perception and video input with a concrete mask output, not just an image embedding.
Limitations
- Segmentation depends on supplied prompts or the documented automatic-mask procedure; output is a mask, not generated video content.
Sources and related Studio records
Sources
Connections
- In areaComputational Models
- Supports functionSegment images and video
- ArtifactSAM 2.1 Hiera Large model files