Sequence-to-Sequence Pre-training with Unified Modality Masking for Visual Document Understanding

Open in new window