Docmatix - a huge dataset for Document Visual Question Answering

Docmatix is a newly released dataset specifically designed for Document Visual Question Answering (DVQA) tasks, comprising over 1 million annotated document images and corresponding questions. It includes diverse document types and complex visual layouts, facilitating the training and evaluation of models on both visual and textual comprehension. This dataset is significant for practitioners as it enables the development of more robust DVQA systems by providing a comprehensive benchmark for model performance and generalization across various document formats.