Image Processing Pipeline

This example demonstrates URL data sources and basic image processing with OpenCV.

What This Example Shows

1. URL as Data Source - The source image is loaded directly from a URL (Wikipedia's Lenna test image)
2. Automatic Caching - URL resources are downloaded and cached in .loom-url-cache/
3. Branching Pipeline - Grayscale output feeds into two parallel processing steps
4. Image Thumbnails - The visual editor shows thumbnails for each image

Pipeline Structure

[URL: Lena image]
        |
        v
   [grayscale]
     /      \
    v        v
[edge_detect]  [blur]
    |            |
    v            v
edges.png   blurred.png

Running the Example

# Navigate to the example directory
cd examples/image-processing

# Run the pipeline
loom examples/image-processing/pipeline.yml

# Or run the full pipeline
loom examples/image-processing/pipeline.yml --all

First run will download the image from Wikipedia and cache it locally. Subsequent runs use the cached version.

Tasks

Task	Description	Input	Output
grayscale.py	Convert to grayscale	Color image	Grayscale image
edge_detect.py	Canny edge detection	Grayscale image	Edge image
blur.py	Gaussian blur	Any image	Blurred image

Parameters

Parameter	Default	Description
blur_radius	15	Blur kernel size (pixels)
edge_threshold_low	50	Canny low threshold
edge_threshold_high	150	Canny high threshold

Clearing the URL Cache

To re-download the source image:

loom clean pipeline.yml

This removes both the generated outputs and the URL cache directory.

Dependencies

• OpenCV (opencv-python-headless or opencv-python)

Install with:

pip install opencv-python-headless

Or install all example dependencies:

pip install -e ".[examples]"