Arabic End-to-End Structured OCR for textbooks

This is the official demo for the Arabic Nougat models. It is an end-to-end Markdown Extraction model that extracts text from images or PDFs and write them in Markdown.

There are three models available:

• arabic-small-nougat: A small model that is faster but less accurate (a finetune from facebook/nougat-small).
• arabic-base-nougat: A base model that is more accurate but slower (a finetune from facebook/nougat-base).
• arabic-large-nougat: The largest of the three (Made from scratch using riotu-lab/Aranizer-PBE-86k tokenizer and a larger transformer decoder model).

Disclaimer: These models hallucinate text and are not perfect. They are trained on a mix of synthetic and real data and may not work well on all types of images.