ICDAR 2021 Competition on Document VisualQuestion Answering
This competition addresses the problem of advancing document understanding and question-answering capabilities for researchers and practitioners in computer vision and document analysis, presenting incremental updates with new tasks and datasets.
The paper reports on the ICDAR 2021 competition for Document Visual Question Answering, introducing a new Infographics VQA task with a dataset of over 5,000 images and 30,000 question-answer pairs, where winning methods achieved scores of 0.6120 ANLS for Infographics VQA, 0.7743 ANLSL for Document Collection VQA, and 0.8705 ANLS for Single Document VQA.
In this report we present results of the ICDAR 2021 edition of the Document Visual Question Challenges. This edition complements the previous tasks on Single Document VQA and Document Collection VQA with a newly introduced on Infographics VQA. Infographics VQA is based on a new dataset of more than 5,000 infographics images and 30,000 question-answer pairs. The winner methods have scored 0.6120 ANLS in Infographics VQA task, 0.7743 ANLSL in Document Collection VQA task and 0.8705 ANLS in Single Document VQA. We present a summary of the datasets used for each task, description of each of the submitted methods and the results and analysis of their performance. A summary of the progress made on Single Document VQA since the first edition of the DocVQA 2020 challenge is also presented.