Bretina - Python Image Testing Framework

Bretina is a python package designed to support testing of the images.

Bretina is designed as an extension of the python unit test module. Provides set of assertions which can be used to verify correctness of the image.

In typical application, content of the device LCD display is captured with a camera and Bretina is used to verify correctness of the defined regions in the image - such as region color, presence of an icon, correctness of the text representation and other aspects.

Documentation

can be found on https://docs.benderrobotics.com/bretina (Not publicly available ATM).

Download and install latest release

Bretina can be downloaded from within the internal BR network by calling:

    $ pip install bretina

For optical character recognition Bretina offers two engines:

• Tesseract
• PaddleOCR

To change the OCR engine set the bretina.OCR_ENGINE to either OcrEngine.TESSERACT or OcrEngine.PADDLE. By default OcrEngine.TESSERACT is set.

PaddleOCR local models

By default PaddleOCR downloads its models from the internet on first use. To use models bundled locally instead, install the models extra. The bretina-models package is hosted on the Bender Robotics private PyPI, so it has to be passed as the extra index:

    $ pip install bretina[models] --extra-index-url https://pypi.benderrobotics.com

When the bretina-models package is available, PaddleOCR automatically uses the local models instead of downloading them.

PaddleOCR on Windows

PaddleOCR's native library (paddlepaddle) requires the Microsoft Visual C++ Redistributable (x64). Without it, the first Paddle OCR call fails with ImportError: DLL load failed while importing libpaddle: The specified module could not be found.. Install the latest redistributable from https://learn.microsoft.com/en-us/cpp/windows/latest-supported-vc-redist?view=msvc-170#latest-supported-redistributable-version and reboot.

Tesseract usage

To utilize Tesseract, it has to be installed as a standalone application and registered into system PATH. Installation files can be downloaded from https://github.com/tesseract-ocr/tesseract (tested with Tesseract version 5). Windows installer is provided by Mannheim University Library at https://github.com/UB-Mannheim/tesseract/wiki.

After the installation, add path to the tesseract.exe to your system PATH.

For the best OCR performance install the slower, but more accurate datasets tessdata_best (https://github.com/tesseract-ocr/tessdata_best). Extract the downloaded archive into the installation directory of the tesseract OCR.

This is an expected structure of the tesseract installation directory:

• C:\Program Files\Tesseract-OCR- tesseract installation
  • \tessdata - original tessdata dataset
    • afr.traineddata
    • ...
  • \tessdata_best - extracted best dataset
    • afr.traineddata
    • ...

Working with the source

Clone the repository first, then check out devel for the latest version, or a feature/* branch for the cutting edge one.

    $ git checkout devel

Development setup

Install the package in editable mode with the development dependencies:

    $ pip install -e .[dev]

Do this before anything else. The version is derived from the git tags by setuptools-scm, which writes the generated src/bretina/__version__.py at install and build time. A freshly cloned, never installed checkout has no such file, so import bretina fails until you install or build once.

Install the git hooks as well, so the formatting and lint gate runs before every commit instead of failing in CI:

    $ pre-commit install

Building the wheel

    $ python -m build

The wheel and the source distribution are written to dist/.

VS Code tasks

The recommended entry point for everyday work are the tasks in .vscode/tasks.json, run through Terminal > Run Task.... They are the same commands CI runs, so a green task means a green pipeline (usually):

Neither ruff nor mypy takes a path argument, both are configured in pyproject.toml.