AdvancedLiterateMachinery

Why we included this project

Alibaba's DuGuang OCR team maintains this research collection for anyone who needs to extract text from messy images and documents. It brings together code and papers covering scene text detection and recognition, document parsing, table recognition, and key-information extraction, along with the CC-OCR benchmark for testing how well large multimodal models actually read. That matters because production OCR usually means choosing between a focused specialist and a generalist vision-language model for each job. The Platypus and OmniParser releases demonstrate a single architecture handling multiple reading tasks, which is useful if you want to reduce the number of separate OCR systems you maintain. Treat it as a research codebase to adapt, not a turnkey product.