Segment Anything（SAM）
Product information, use cases, and access for Segment Anything（SAM）.
Pricing information
No verified public pricing is available yet.
DevPrice organizes public information and does not sell Segment Anything（SAM） subscriptions. Prices and availability are determined by Segment Anything（SAM）.
What is Segment Anything（SAM）?
Segment Anything (SAM) is an image segmentation model from Meta AI Research for work that requires separating objects or regions within an image. It uses prompts such as points or bounding boxes to produce object masks, and it can also generate masks for multiple objects in an image. The model is suited to developers, researchers, and visual workflows that need object boundaries as an output.
A user indicates the area of interest with a supported prompt, and SAM produces a corresponding segmentation mask. It is also described as supporting zero-shot performance across image segmentation tasks, so it can be explored without building a separate task-specific model first. The resulting masks can serve as a starting point for later image analysis, editing, measurement, or other processing that depends on distinct object regions.
Key features of Segment Anything（SAM）
Point-Prompt Segmentation
Uses a user-provided point as a visual prompt to identify a relevant object and produce its mask, supporting segmentation workflows that begin with a precise location in the image.
Box-Prompt Segmentation
Accepts a bounding-box prompt to define the approximate target area, then generates an object mask from that region for workflows that start with a coarse spatial indication.
Multi-Object Masking
Can generate masks for multiple objects in an image, helping users separate distinct visual elements from the overall scene for subsequent analysis or image editing.
Zero-Shot Segmentation
Is described as providing zero-shot performance across image segmentation tasks, allowing users to explore object masks without first training a separate model for one narrowly defined task.