DocVQA: A Dataset for VQA on Document Images
This provides a new benchmark for VQA on documents, addressing a domain-specific need for improved document analysis, but it is incremental as it builds on existing VQA and reading comprehension datasets.
The authors introduced DocVQA, a dataset of 50,000 questions on over 12,000 document images for visual question answering, and reported baseline models achieving lower accuracy than human performance (94.36%), highlighting a gap in understanding document structure.
We present a new dataset for Visual Question Answering (VQA) on document images called DocVQA. The dataset consists of 50,000 questions defined on 12,000+ document images. Detailed analysis of the dataset in comparison with similar datasets for VQA and reading comprehension is presented. We report several baseline results by adopting existing VQA and reading comprehension models. Although the existing models perform reasonably well on certain types of questions, there is large performance gap compared to human performance (94.36% accuracy). The models need to improve specifically on questions where understanding structure of the document is crucial. The dataset, code and leaderboard are available at docvqa.org