Annotations: Segment Anything

experimental

Run Segment Anything mask inference in-browser (or on a configured GPU server) as an annotation tool, and expose it as a local segmentation driver for the pathology foundation-model namespace. Requires the Annotations plugin.

ID	sam-segment-tool-experimental
Version	1.1.0
Author	Mia
Categories	Annotations
Source	plugins/sam-segment-tool-experimental

Keywords: segment · sam · mask · ai · roi

Dependencies

• Annotations Module
• Pathology Foundation Models

Required configuration

• servers
• models

Additional configuration keys

models, transformers, servers

See include.json for details and defaults.

Documentation

Annotations: Segment Anything (experimental)

Runs Segment Anything mask inference in-browser (via @huggingface/transformers) as an annotation tool, and exposes it as a local segmentation driver for the pathology foundation-model namespace. Requires the Annotations plugin.

Configuration

Deployment config lives under this plugin's ENV.plugins["sam-segment-tool-experimental"] block (merged into include.json).

models

Map of model id → HuggingFace repo used for in-client execution:

"models": {
  "slimsam-77-uniform": "Xenova/slimsam-77-uniform",
  "medsam-vit-base": "Xenova/medsam-vit-base"
}

servers

By default only in-client execution is supported, so servers is an empty array. To offload inference to a GPU server hosting supported segmentations for transformers, add entries of the form:

"servers": [
  { "name": "A10", "path": "https://example.com/slimsam-a10" }
]

include.json is strict JSON (no comments) — it is parsed with JSON.parse by the module/plugin loader. Keep configuration examples here in the README rather than as inline comments in the manifest.