Pretrained Document Attention Networks on two datasets: READ 2016 and RIMES 2009. "fcn_line_syn" indicates the pretraining stage during which the encoder is trained on line-level synthetic images "dan" corresponds to the main training on document-level images. Denis Coquenet, Clément Chatelain, Thierry Paquet, "DAN: a Segmentation-free Document Attention Network for Handwritten Document Recognition", Transactions on Pattern Analysis and Machine Intelligence, 2023 Source code: Github Paper: Arxiv

{"references": ["https://arxiv.org/abs/2203.12273", "https://github.com/FactoDeepLearning/DAN"]}