imgparallel

A toolbox for pre-processing and formatting image datasets.

imgparallel is a command line interface and python library for processing image datasets stored as individual files. Currently the tool exists as an MVP, supporting very minimal transforms and limited flexibility.

Installation

From PyPI

pip install imgparallel

From Source

git clone https://github.com/jackson-waschura/imgparallel.git
cd imgparallel/
pip install -e .

Examples

Here is an example usage which resizes all images from test_data/input_images to 512x512 then saves them to test_data/output_images as jpegs.

imgparallel --src test_data/input_images --dst test_data/output_images --resize 512x512 --format jpg --proc_per_stage 1

The CLI can also be invoked like this:

python3 imgparallel/cli.py --src test_data/input_images --dst test_data/output_images --resize 512x512 --format jpg --proc_per_stage 1

To Do

• Implement a simple MVP CLI tool for reformatting an image dataset (image format, resize)
• Evaluate the output of running the tool
• Implement sanity checks / tests for permissions before running the pipeline.
• Add an alias / script into the package so it can be run by name from anywhere
• Implement tests to evaluate correctness of the tool
• Use wheel and twine to publish the MVP tool
• Draft a design for more complex transforms (multiple resolution outputs, cropping, etc)