Cellpose Tools

Configuration File: cellpose_tools.json Tool Type: Local Tools Count: 1

This page contains all tools defined in the cellpose_tools.json configuration file.

Available Tools

Cellpose_segment_image (Type: CellposeTool)

Segment cells or nuclei in a single microscopy image LOCALLY using the deep-learning Cellpose mod…

Cellpose_segment_image tool specification

Tool Information:

  • Name: Cellpose_segment_image

  • Type: CellposeTool

  • Description: Segment cells or nuclei in a single microscopy image LOCALLY using the deep-learning Cellpose model (no API, no key). Given a local image file path, returns the number of segmented objects, per-object area (pixels) and centroid (y,x), the image shape, and optionally the path to a saved integer label-mask image. Choose ‘model_type’ = ‘cyto3’/’cyto’/’cyto2’ for whole-cell/cytoplasm segmentation or ‘nuclei’ for nuclei. NOTE: cellpose 3.x honors ‘model_type’; cellpose 4.x uses a single unified model (CPSAM) and ignores ‘model_type’ (the ‘model_used’ field and a ‘note’ report what actually ran). Requires the optional ‘cellpose’ package (pip install cellpose; pulls in torch). On first use cellpose downloads model weights (small in 3.x, ~1 GB CPSAM in 4.x) and caches them, so the first call is slow; later calls reuse cached weights. Returns a clean error if cellpose is not installed or the image cannot be read.

Parameters:

  • image_path (string) (required) Path to a local microscopy image file (.tif, .tiff, .png, .jpg, .jpeg, .bmp).

  • model_type (string) (optional) Cellpose model to use. ‘cyto3’/’cyto’/’cyto2’ for cells/cytoplasm, ‘nuclei’ for nuclei. Honored on cellpose 3.x; ignored by the unified model on 4.x. Default ‘cyto3’.

  • diameter (number) (optional) Expected object diameter in pixels. Omit or 0 to let cellpose estimate it automatically.

  • channels (array) (optional) Two-element [cytoplasm, nucleus] channel spec. Use [0,0] for a grayscale image (default). Used by cellpose 3.x; 4.x infers channels.

  • save_mask (boolean) (optional) If true, save the integer label mask as a 16-bit PNG and return its path. Default false.

  • mask_output_path (string) (optional) Where to write the mask when save_mask is true. Defaults to ‘<image_path>_cp_masks.png’.

Example Usage:

query = {
    "name": "Cellpose_segment_image",
    "arguments": {
        "image_path": "example_value"
    }
}
result = tu.run(query)