Press Releases > OCR Wins ICDAR2019 Receipt Recognition Competition

OCR Wins ICDAR2019 Receipt Recognition Competition

May 30, 2019

At ICDAR 2019, the joint innovation laboratory formed by Huazhong University of Science and Technology (HUST) and HUAWEI CLOUD achieved 96.43% precision in the Scanned Receipt OCR competition, ranking first in the world.


End-to-end text recognition requires the integration of detection and recognition and is the most important of all OCR tasks. ICDAR Competition Link http://rrc.cvc.uab.es/?ch=13

The International Conference on Document Analysis and Recognition (ICDAR) is the most authoritative international conference in the field of document analysis, as well as one of the most authoritative competitions in the global OCR field. It is held every second year and this year's was the 15th.

The application of SROIE technology is highly important in the fields of finance, accounting and taxation, greatly simplifying the process of document-intensive work and facilitating the realization of office automation. Compared with conventional OCR tasks, scanned receipts OCR and information extraction (SROIE) faces challenges such as low receipt quality, highly diverse layouts, and higher accuracy requirements.

In order to overcome these challenges, HUAWEI CLOUD and HUST 's algorithmic technicians designed a new network model and algorithm based on deep learning, with the aid of the super-powerful computing power and fast iterative algorithms of the HUAWEI CLOUD platform. Through the integration of multiple model structures and continuous optimization of training strategies, a more efficient and accurate recognition model was obtained. The Huawei & HUST Smart Innovation Lab was established in January this year. The winning of this competition demonstrates that the research cooperation between these two parties on OCR and other fields has achieved initial success.


HUAWEI CLOUD+AI's next-generation OCR cloud service

OCR receipt recognition is applied extensively in fields such as finance, insurance, medicine and financial reimbursement. To take an insurance company as an example, in the traditional claims process, the procedures of adjustment, compensation, payment, and filing are all done manually, requiring 5 to 6 days of manual work. With the HUAWEI CLOUD OCR service, the entire claims process can be completed automatically in about 15 minutes. At the same time, the claims system can work 24 hours a day, year-round, without the need for manual collaboration between multiple positions, saving massive labor costs and greatly improving the efficiency of claims.

Using innovation and technology leadership, on May 26th, “HUAWEI CLOUD’s Smart OCR Data Product” won the Big Data Expo’s“2019 Leading Science and Technology Achievement Award – New Product Award”.

At present, HUAWEI CLOUD's OCR products have been applied with success globally in the fields of medicine, customs, logistics, finance, insurance, government, transportation, automobiles, manufacturing and other business areas with cross-system information integration needs, helping industries automate data collection and digital information management, improve efficiency and reduce costs, and achieve a high degree of customer recognition.