Request: ability to return character bounding boxes from generator

[notmatthancock]

For over-segmentation style OCR approaches, it would be helpful to have character bounding boxes. Glancing at the source for computer_text_generator.py it seems feasible with the current implementation. Although, I'm not accounting for potential skews and other geometric transformations that might add considerable more difficulty.

[Belval]

The PR is outdated, but the feature was implemented just never merged: https://github.com/Belval/TextRecognitionDataGenerator/pull/144

I could try to fix and merge it.

[alikaz3mi]

Hello. Has this feature merged to the pip installation yet?