Engineers from Infrrd's Research team emerged as the winners of the 2021 edition of the DocVQA challenge (Document Collection DocVQA task) organized by ICDAR 2021.
San Jose, CA - ICDAR 2021, a premier international event for scientists and practitioners involved in document analysis and recognition held its 2021 Edition of the DocVQA (Document Visual Question Answering) challenge earlier this year. In the Document Collection VQA task of the challenge, California-based Infrrd’s team of engineers made it to the top of the leaderboard and won the challenge with their ingenious approach.
ICDAR’s DocVQA challenge was created with a sophisticated vision. Contemporary Document Analysis and Recognition (DAR) research, which tends to focus on generic information extraction tasks, fails to provide human-ready insights from the extracted data. The DocVQA challenge tasks inspired participants to take this a step ahead where the document content is extracted and used to answer high-level tasks defined by the human consumers of this information.
Infrrd’s team of engineers from their AI research lab solved this task with the Infrrd-RADAR (Retrieval of Answers by Document Analysis and Re-ranking) method. The solution method authored by JiangLong He, Aditya Kumar Sarda, Deepak Kumar, Cesar Duran was able to solve the DocVQA challenge task in the most effective and efficient way. Infrrd’s team beat PingAn OneConnect GammaLab, one of China’s top FinTech companies, iFLYTEK research and engineers from the tech giant Amazon to race up the leaderboard.
The team was honored to present their method at the 16th International Conference on Document Analysis and Recognition in Lausanne, Switzerland held on September 5-10, 2021.
Read more about the ICDAR’s DocVQA challenge and the results here.
ABOUT INFRRD
Infrrd, headquartered in San Jose, California, is driving innovation ahead in the AI space with its proprietary Intelligent Document Processing (IDP) platform. Our platform allows enterprises to classify and extract data from semi-structured and unstructured documents. It offers pre-built, out-of-the-box models for use cases such as invoices, claims, and bank statements for easy adoption, while also allowing enterprises to create and train custom models. The platform has advanced cognitive capabilities, a user-friendly interface, and the flexibility to process different data types, contributing to its rapid growth in the market.
ABOUT ICDAR
ICDAR is the premier international event for scientists and practitioners involved in document analysis and recognition, a field of growing importance in the current age of digital transition. In 2021, the 16th edition of this flagship conference will be held for the first time in Switzerland. Read more about the conference here.
