Documentation for internal methods?

[tarrinw]

We are working on an application using Kraken for both training and recognition. The command-line tools work well, but it would be easier to deploy in a dynamic python envoronment (e.g. Jupyter / Google colab) if we could access the internal / API methods directly. Presumably this is just a matter of documentation?

[tarrinw]

Actually one can use the kraken shell commands (prefixed with '!') in these environments, but it would still be useful and slightly more efficient when building a wrapper application in Python not to have to call the shell which then runs code in Python anyway...

[johnlockejrr]

I think that's why the doc folder is there... https://github.com/mittagessen/kraken/tree/main/docs https://github.com/mittagessen/kraken/blob/main/docs/api_docs.rst https://github.com/mittagessen/kraken/blob/main/docs/api.rst Quote:

from kraken import pageseg

seg = pageseg.segment(bw_im)
seg

Segmentation(type='bbox',
             imagename='foo.png',
             text_direction='horizontal-lr',
             script_detection=False,
             lines=[BBoxLine(id='0ce11ad6-1f3b-4f7d-a8c8-0178e411df69',
                             bbox=[74, 61, 136, 101],
                             text=None,
                             base_dir=None,
                             type='bbox',
                             imagename=None,
                             tags=None,
                             split=None,
                             regions=None,
                             text_direction='horizontal-lr'),
                    BBoxLine(id='c4a751dc-6731-4eea-a287-d4b57683f5b0', ...),
                    ....],
             regions={},
             line_orders=[])

You don't have to build a wrapper application in Python because kraken is written in Python and can be imported.

[mittagessen]

On 24/04/22 12:32AM, Tarrin Wills wrote:

We are working on an application using Kraken for both training and recognition. The command-line tools work well, but it would be easier to deploy in a dynamic python envoronment (e.g. Jupyter / Google colab) if we could access the internal / API methods directly. Presumably this is just a matter of documentation?

The website has both an API tutorial and the API reference. For some more examples on how to use the API for inference the scripts in contrib/ can be a good starting point as they tend to have less boiler plate than the CLI code.