DocMAE: Document Image Rectification via Self-supervised Representation Learning | IEEE Conference Publication | IEEE Xplore