anybioimage

name

anybioimage

description

Integrate and use the anybioimage BioImageViewer widget to display biological images, overlays, and annotations in Jupyter and marimo notebooks. Use this skill when asked to display images with BioImageViewer, add mask overlays or segmentation results, annotate images with rectangles/polygons/points, enable SAM (Segment Anything Model) for interactive segmentation, load multi-dimensional 5D images (TCZYX) with BioImage, show multichannel fluorescence images with per-channel colors and contrast, export annotations as DataFrames, or integrate anybioimage into a marimo or Jupyter workflow.

anybioimage Skill

anybioimage provides BioImageViewer, an interactive widget for visualizing biological images with mask overlays, annotations, and SAM segmentation in Jupyter and marimo notebooks.

Installation

uv pip install -e ".[all]"   # Full install with all dependencies
uv pip install -e ".[sam]"   # SAM support only

Quickstart

Jupyter

from anybioimage import BioImageViewer
import numpy as np

viewer = BioImageViewer()
viewer.set_image(np.random.randint(0, 255, (512, 512), dtype=np.uint8))
viewer  # Display inline

Marimo

import marimo as mo
from anybioimage import BioImageViewer

@app.cell
def _():
    viewer = BioImageViewer()
    viewer.set_image(image_data)
    return mo.ui.anywidget(viewer)

Marimo rules:

• Always wrap with mo.ui.anywidget(viewer)
• Only edit code inside @app.cell decorators
• Run marimo check --fix after editing

Image Input

Input	Method	Notes
numpy array	set_image(arr)	Auto-squeezed to 2D
BioImage	set_image(BioImage("file.tif"))	Full 5D, lazy loading
OME-Zarr	set_image(BioImage("file.ome.zarr"))	Multi-resolution

BioImage (5D multichannel)

from bioio import BioImage

viewer = BioImageViewer()
img = BioImage("multichannel.ome.tiff")
viewer.set_image(img)
# Dimension controls (T, C, Z) appear automatically in the UI

Masks / Overlays

# Add a segmentation mask
mask_id = viewer.add_mask(labels_array, name="Nuclei", color="#ff0000", opacity=0.5)

# Contour-only overlay
mask_id2 = viewer.add_mask(cell_labels, name="Cells", contours_only=True, contour_width=2)

# Convenience: clear all + add one
viewer.set_mask(labels_array, name="Segmentation")

# Update settings
viewer.update_mask_settings(mask_id, opacity=0.8, visible=False)

# Remove
viewer.remove_mask(mask_id)
viewer.clear_masks()

# Inspect
print(viewer.masks_df)  # DataFrame: id, name, visible, opacity, color, contours

color accepts hex strings. Colors auto-cycle from a 10-color palette if not specified.

Annotations

# Access as DataFrames (user-drawn or programmatically set)
rois     = viewer.rois_df      # columns: id, x, y, width, height
polygons = viewer.polygons_df  # columns: id, points (list of {x, y})
points   = viewer.points_df    # columns: id, x, y

# Export
rois.to_csv("rois.csv")

# Set programmatically
viewer.rois_df = existing_rois_df

# Clear
viewer.clear_rois()
viewer.clear_polygons()
viewer.clear_points()
viewer.clear_all_annotations()

Drawing tools (keyboard shortcuts): Pan (P), Select (V), Rectangle (R), Polygon (G), Point (O)

SAM Integration

# Enable (downloads model on first use via Ultralytics)
viewer.enable_sam("mobile_sam")   # Recommended: ~40MB, fast

# Draw rectangles or click points → SAM auto-segments
# SAM results appear as a mask overlay layer

viewer.clear_sam_masks()  # Clear SAM results
viewer.disable_sam()      # Disable and unload model

Model options: "mobile_sam" (default), "sam_b", "sam_l", "fast_sam"

Visual Settings

viewer.image_visible = True
viewer.image_brightness = 0.2   # -1.0 to 1.0
viewer.image_contrast = 0.3     # -1.0 to 1.0
viewer.canvas_height = 600      # pixels

# Annotation colors
viewer.roi_color = "#ff0000"
viewer.polygon_color = "#00ff00"
viewer.point_color = "#0000ff"
viewer.point_radius = 5         # pixels

Reference Files

• api_reference.md — Full method signatures, all traitlets, channel settings, internal details
• examples.md — Complete workflow examples (fluorescence imaging, segmentation pipelines, annotation export)