Add object masks to images without retraining.
Guides implementation of image segmentation with SAM, producing object masks from prompts or automatic mask generation.
When to use it
Use when implementing image segmentation without task-specific training, interactive annotation, training-data generation, or segmentation pipelines.
It changes how the agent implements SAM-based image segmentation in project code.
What you provide
The guide centers on the segment-anything library, which supplies the model registry and predictor.
Declares torch version 1.7.0 or newer as a dependency.
This skill
dl.fbaipublicfiles.com
Downloads SAM model checkpoints
segment-anything
Installs SAM from GitHub