Example of the “Historical Map Text Reading” task at ICDAR 2025 (Image courtesy of Naver).
SEOUL, Oct. 8 (Korea Bizwire) — Naver Cloud has claimed first place at a leading international competition for optical character recognition (OCR), cementing its standing as a global leader in text recognition technology.
According to the ICT industry on Wednesday, Naver Cloud ranked first at the OCR competition held during the International Conference on Document Analysis and Recognition (ICDAR) 2025, organized by the International Association for Pattern Recognition (IAPR) from September 16 to 21 in China.
The ICDAR competition is one of the world’s most prestigious contests in document analysis and recognition, first established in 1991.
This marks the fourth time Naver and its cloud affiliate have won the global OCR title, following victories in 2018, 2019, and 2022. Naver Cloud took the top spot in the “Historical Map Text Reading” category, which challenges participants to detect and interpret text from old maps containing complex layouts and distorted or overlapping lettering.
The company’s end-to-end model, capable of simultaneously extracting text and spatial coordinates from document images, was praised for its accuracy. Naver Cloud outperformed competitors such as the Swiss Federal Institute of Technology in Lausanne (EPFL) to secure the top position.
OCR has become an increasingly vital component of multimodal large language models (LLMs), which combine visual and linguistic understanding.
A recent Naver Cloud paper presented at the North American Chapter of the Association for Computational Linguistics (NAACL) showed that LLMs without strong OCR capabilities often misinterpret or skip image-based text, leading to faulty reasoning. Models with dedicated or embedded OCR systems, by contrast, achieved markedly higher accuracy.
Global competition in the field is intensifying, with France’s Mistral AI releasing a high-precision “Mistral OCR” earlier this year that can process handwritten and printed text.
Naver Cloud plans to integrate its award-winning OCR technology into its Clova OCR platform starting next year. “We aim to introduce a next-generation OCR solution capable of reading unstructured documents and curved text layouts with high precision,” a company official said, adding that the firm intends to expand its expertise into broader vision AI technologies.
Kevin Lee (kevinlee@koreabizwire.com)
