Going Full-TILT Boogie on Document Understanding with Text-Image-Layout Transformer

Benchmark Model Rank Results
document-image-classification-on-rvl-cdipTILT-Large#7Accuracy: 95.52%
document-image-classification-on-rvl-cdipTILT-Base#12Accuracy: 95.25%
visual-question-answering-vqa-on-infographicvqaTILT-Large#16ANLS: 61.20
visual-question-answering-on-docvqa-testTILT-Large#56ANLS: 0.8705
visual-question-answering-on-docvqa-testTILT-Base#63ANLS: 0.8392