Note Dhivehi OCR: in the code switch the BIDI did not work well and the made images are wrong.

Note syntactic, dhivehi OCR, no bidi, wrong images in case of code switch otherwise fine.

Note real scene, chinese, multiple dataset

Note arabic pages with header and footer

Note 1% sampled from https://huggingface.co/datasets/linxy/LaTeX_OCR

Note Vietnam Finance PDF reports with OCR data

Note ocr tables

Note english pages without noise. not clear from where the annotations come from. easy.

Note very good receipt and invoices

Note malayalam, word level

Note persian syntatic

Note Nastaliq collection urdu

Note Easy english word

Note Old Vietnam in chinese characters

Note ❌ no image