These samples all use Juypter Notebooks.
Python requirements:
pip install -r requirements.txt
- Mindee is used for the OCR data extraction
- JSON to read the outputs
- cv2, pyplot numpy are all used to create the charts, and draw shapes on the charts
- Math is used to round the pixel math to an integer.
- Pretty prints the JSON nicely.
To start a Jupyter Notebook:
- install Jupyter
- Start the Jupyter instance
jupyter notebook - A tab will open in the browser and you can naviagte to the ipynb Jupyter file you wish to examine.
To use these dociuments, you'll need a Mindee API key. Each APi endpoint uses a different key - so you must generate the key for the endpoint you are working with.
There are 4 docs:
- invoice
- receipt
- passport
- license plates
Additionally, there are sample images for all of the endpoints. To test with a new document, sinply add to the directory, and change the path for the 'image' variable.
In the invoice and license plate examples - not only are the various values extracted from the document, but a box is drawn around the location of several of the attributes. This is done by retrieving the polygon data from the Mindee response, converting to a pixel corrdiante, and the using openCV to draw a rectangle on the image.