arjunrajlaboratory/nimbusimage/images

Claude Code and agent skill by arjunrajlaboratory/nimbusimage

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Fetch image data from NimbusImage datasets as numpy arrays using the nimbusimage Python API. Use this skill when the user wants to retrieve image frames, composites, z-stacks, crops, or channel data from a NimbusImage dataset. Also use when you see ds.images in code, or the user mentions getting pixel data, viewing channels, making composites, or processing image arrays from their imaging server.

Install

npx vskill@latest install arjunrajlaboratory/nimbusimage/images

SKILL.md

NimbusImage — Images

The ds.images accessor retrieves image data as numpy arrays from a NimbusImage dataset. Images are fetched from the Girder large_image backend via a pickle-encoded region API.

Single frame

python
import numpy as np
# Get one frame as a 2D numpy array (H, W)
img = ds.images.get(channel=0, z=0, time=0)
print(f"Shape: {img.shape}, dtype: {img.dtype}")
# Specify XY position (for multi-position datasets)
img = ds.images.get(channel=0, z=0, time=0, xy=1)

Cropping

python
# Crop to a region: (left, top, right, bottom) in pixel coordinates
crop = ds.images.get(channel=0, z=0, time=0, crop=(100, 200, 500, 600))

All channels

python
# Get all channels at one location as a list of 2D arrays
channels = ds.images.get_all_channels(z=0, time=0)
for i, ch in enumerate(channels):
    print(f"Channel {i}: shape={ch.shape}")

Z-stacks and time series

python
# Z-stack as a 3D array (N_z, H, W)
z_stack = ds.images.get_stack(channel=0, axis="z")
# Time series as a 3D array (N_t, H, W)
time_series = ds.images.get_stack(channel=0, axis="time")
# At specific location
z_stack = ds.images.get_stack(channel=0, axis="z", xy=0, time=0)

Composite image

Creates an RGB composite merging visible channels using layer settings (contrast, pseudocolor) from the dataset's collection configuration:

python
# Composite as (H, W, 3) numpy array
rgb = ds.images.get_composite(dtype="uint8")   # [0, 255]
rgb_f = ds.images.get_composite(dtype="float64")  # [0.0, 1.0]
# At specific location
rgb = ds.images.get_composite(z=3, time=0, dtype="uint8")

The composite is drawn the way the app draws it for you: layer settings from ds.collections.layers (your own view's configuration) — channel visibility, pseudocolor, contrast with your per-view overrides (percentile blackPoint/whitePoint are percents of the frame histogram's min..max, not quantiles), and each layer's z/time/xy slice setting, including max projections. It blends with "lighten" mode by default. Pass max_size=1024 for large slides (downsampled on the server) and crop=(left, top, right, bottom) to zoom in.

Iterating all frames

python
# Iterate over every frame in the dataset
for frame_info, img in ds.images.iter_frames():
    print(f"Frame {frame_info.index}: ch={frame_info.channel}, "
          f"z={frame_info.z}, t={frame_info.time}, shape={img.shape}")

Writing images (worker context)

For Docker workers that need to write processed images back:

python
# Requires pip install nimbusimage[worker]
with ds.images.new_writer() as writer:
    for frame_info, img in ds.images.iter_frames():
        processed = some_processing(img)
        writer.add_frame(processed, **frame_info.to_large_image_params())
# Automatically writes TIFF and uploads to dataset folder on context exit

Common patterns

Save a frame as PNG
python
from PIL import Image
img = ds.images.get(channel=0, z=0, time=0)
# Normalize to 0-255 if needed
if img.dtype != np.uint8:
    img = ((img - img.min()) / (img.max() - img.min()) * 255).astype(np.uint8)
Image.fromarray(img).save("frame.png")
Save composite as PNG
python
from PIL import Image
rgb = ds.images.get_composite(dtype="uint8")
Image.fromarray(rgb).save("composite.png")
Maximum intensity projection
python
z_stack = ds.images.get_stack(channel=0, axis="z")
mip = z_stack.max(axis=0)  # (H, W)

Key details

  • All image methods return numpy arrays (squeezed to 2D for single frames).
  • The dataset's dtype property tells you the source data type (e.g., uint8, uint16).
  • ds.shape returns (height, width) — numpy convention, not (width, height).
  • Composite requires layer settings in the dataset's collection. If no layers are configured, it returns a blank image.
  • Frame coordinates: channel, z, time, xy are all 0-indexed integers.

For complete ImageAccessor signatures, read references/api-overview.md. For coordinate and composite caveats, read references/gotchas.md.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

See the full security report

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

See evaluation results

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/arjunrajlaboratory/nimbusimage/images/badge)](https://verified-skill.com/skills/arjunrajlaboratory/nimbusimage/images)

Something wrong with this skill or its page? Tell us.