Cache images and crop

Closes #150 (closed)

changed milestone to %ML Prod - Summer 2023

added P1 label

assigned to @yschneider

added 1 commit

• 15355f11 - do not change test fixture

Compare with previous version

I took inspiration from what I did on atr/data-generator. We could support more parameters to support all extraction modes like we do over there. This makes the data generation process much faster because we download way less images.

requested review from @mboillet

In the future, we should probably rename the teklia_line_image_extractor repo since it can be used to extract any type of subelements like paragraphs, articles, acts..., and not just lines.

added 1 commit

• 31564c75 - clearer docstring

Compare with previous version

resolved all threads

approved this merge request

merged