1 paper
Yuliang Liu, Zhang Li, Ziyang Zhang +11
Mainstream visual encoders are pretrained on natural images and cannot be effectively applied to document images without document-oriented adaptation, as dense text and fine-graine…