Search
Recently, Baidu officially made its end-to-end OCR model, Unlimited OCR, available as open-source software. Just one day after its release, the model rapidly ascended to the pinnacle of GitHub’s Daily Trending and Python charts. Simultaneously, it claimed the number one spot on both HuggingFace’s Global Model Overall Trending chart and the Multimodal Model Trending chart.
Specifically tailored for parsing lengthy documents, Unlimited OCR boasts a total parameter size of 3 billion, with only approximately 570 million parameters activated during the inference process. Public evaluations reveal that the model achieved an impressive comprehensive score of 93.92% on the OmniDocBench v1.6 benchmark test, setting a new benchmark for end-to-end OCR performance.
-
C114 Communication Network -
Communication Home
