Hikvision’s Deep Learning-Based Word Recognition Technology Wins 1st Prize in ICDAR 2016 Robust Reading Competition

Global

Global - English

Asia

Europe

Oceania

ANZ - English

Latin America

North America

Middle East and Africa

Search news

Cancel

Reset

Submit

April 12, 2017 Hikvision, the world’s leading supplier of innovative video security products and solutions, is proud to announce that its Optical Character Recognition (OCR) Technology based on Deep Learning and led by the company’s Research Institute has won the first price in the ICDAR 2016 Robust Reading Competition. The Hikvision team substantially surpassed both strong domestic and foreign competitors in three Word Recognition challenges, including Born-Digital Images, Focused Scene Text, and Incidental Scene Text, demonstrating that the Word Recognition Technology by Hikvision reached the world’s top level.

The International Conference on Document Analysis and Recognition (ICDAR) is held by the International Association for Pattern Recognition (IAPR). This competition is the most influential for OCR technology worldwide. The competition has been held five times since 2003. Since the latest competition in 2015, 2,367 teams from 82 countries have been involved in the competition, including Google, Microsoft, Baidu, Samsung, and Megvii. These competitions have greatly spurred the development of Word Recognition Technology.

The words to be recognized in these challenges are located in complicated image backgrounds. There are multiple adverse factors for imaging, such as noise, blur, transparency, image tilt, and much more. Moreover, different fonts need to be dealt with as well. Hikvision OCR technology has overcome many challenges and won the worldwide first price in the three Word Recognition competitions. Specifically, it surpassed the second place finisher by 5.6% in the Web and Email Born-Digital Images challenge, and by 3.4% and 3.1% respectively in the Focused Scene Text challenge and the Incidental Scene Text challenge.

With the assistance of advanced OCR technology, the Hikvision Word Recognition system is capable of dealing with more complicated Word Recognition scenarios, such as those with stains and blurs, background noise and deformation, extreme weather impact, etc. As of now, Hikvision’s vehicle plate recognition system—based on this new OCR technology—has been applied in over 60locations around the world. Of note are electronic police, entrance/exit control, and parking systems.

In addition, the results of this competition can also be applied in multiple other sectors to greatly improve product performance and application effectiveness. These applications include tag recognition in the machine vision sector, ID and license recognition in traffic and commercial sectors, malicious image recognition over the Internet, and banner and street view recognition in the urban management sector.

Hikvision has seized opportunities to develop Artificial Intelligence (AI), and the company is leveraging its Big Data and superior AI technology to improve its OCR technology performance. Hikvision continually strives to advance machine intelligence and automation, continuously creating value for its customers in areas of video security, intelligent perception, scenario understanding and much more.