How to Use Advanced Smart Text Recognition Technology
To Solve the Challenges of Traditional OCR (Optical Character Recognition) Applications?
At the recent 2022 China Image Graphics Conference forum on “Frontier Technologies and Industrial Applications of OCR”, Dr. Ding Kai, Director of Natural Language Algorithm R&D at Hehe Information Company, introduced the company’s smart text recognition and image processing technology, which was unanimously recognized by experts from top research institutions such as the Chinese Academy of Sciences, Peking University, and Lenovo Research Institute as the “key” to solving these challenges.
Dr. Ding introduced that although OCR technology has developed for over a century, there are still urgent issues to address today, such as severe degradation of document image quality, difficulties in text detection and layout analysis, low recognition rates for text under non-standard conditions, and poor structured intelligent understanding capabilities. In the advancement of OCR technology, enhancing document image quality is an important research direction, which needs to overcome common interference conditions in modern text image processing, such as page curvature, shadow occlusion, moiré patterns, and image blurriness.
Hehe Information Company’s smart text recognition and image processing technology, by incorporating AI (Artificial Intelligence) technology, can help various application fields simplify downstream document processing tasks and improve the efficiency and accuracy of text recognition.
Taking curvature correction as an example, Dr. Ding introduced to the experts at the forum the principles and pros and cons of methods based on text line fitting and coordinate transformation, as well as those based on optimizing text line correction. For these defects, the system architecture based on displacement field network learning adopted by Hehe Information Company can effectively solve various correction problems of curved document images.
At the same time, to better address the issues of complex document layouts, insufficient training samples, and long and inefficient customization tuning cycles in different businesses, Hehe Information Company launched the TextIn Studio smart text recognition training platform, which integrates multiple modules such as underlying resources, data, model training, integrated deployment, and service management applications to specifically solve various problems and establish a closed loop between business processes, achieving automated model training and deployment.
It is reported that TextIn Studio has produced a large number of document digitization models for different scenarios, covering nearly a hundred types of document image preprocessing, text recognition and understanding, and document format conversion services, comprehensively covering document types related to both enterprise and personal work and life. Currently, Hehe Information Company has initiated a limited-time experience activity targeting the needs of college researchers through the TextIn mini-program, where teachers and students at universities can register and bind their “edu” suffix educational email to receive 1 million OCR service requests for free each year.
Additionally, during this year’s China Image Graphics Conference, the award ceremony for the third CSIG Image Graphics Challenge Finals was also held. The CSIG Image Graphics Challenge aims to promote the development and application of image graphics technology and related industries in China, solve technical problems faced by enterprises, and help enterprises attract more outstanding talent.
The team composed of Hehe Information Company and related universities and enterprise ecosystem partners, leveraging algorithm advantages in visual key information understanding and practical experience in multi-language receipt recognition scenarios, not only won the championship in the