[Feature Request] Recognize text styles including size, font type, color, boldness, italics

[atlury]

It would be great if you can look at adding a feature of style recognition and transfer. This along with layout preservation would be a great asset to the OCR pipeline.

[atlury]

Here is one paper/repo "remotely" related to the above

https://github.com/uchidalab/TrueTypeTransformer
https://das2022.univ-lr.fr/wp-content/uploads/OS-slides/72.pdf

[atlury]

Kosmos-2.5: A cutting-edge multimodal literate model revolutionizing text-intensive image understanding. This looks interesting, you can probably explore a bit.

To quote
"Kosmos-2.5 excels in: (1) generating spatially-aware text blocks, where each block of text is assigned its spatial coordinates within the image, and (2) producing structured text output that captures styles and structures into the markdown format. The model can be adapted for any text-intensive image understanding task with different prompts through supervised fine-tuning."