IT Home, October 30: On the 27th of this month, Baidu officially released and open-sourced the ERNIE Unlimited OCR 2 model. It elevates document parsing limits from dozens of pages to “unlimited,” capable of parsing all seven books of the Harry Potter series—totaling over 3,600 pages—in a single operation.
Unlimited OCR 2 natively supports layout element segmentation and recognition, with capabilities for cross-page text merging and table recognition and merging, enabling more, faster, and more cost-effective processing under the same computational power.
Baidu reported that Wenxin Unlimited OCR 2 achieved a score of 84.6 on the MPDocBench benchmark for multi-page document parsing and 77.1 on the MarathonDocBench benchmark, specifically designed to evaluate the parsing capability of ultra-long documents of hundreds of pages, achieving state-of-the-art end-to-end performance in both.
Reference:
https://ai.baidu.com/support/news?action=detail&id=3290