Joint recognition of text and layout in historical Russian documents

In this paper, we evaluated the Document Attention Network (DAN), first end-to-end segmentation-free architecture on Historical Russian Documents. The DAN model jointly recognizes both text and layout from whole documents, it takes documents any size as an input output well logical tokens. For comparison purposes, conduct our experiments Digital Peter dataset has been recognized at line-level. Dataset consists of Great manuscripts; ground truths are represented according to a sophisticated XML schema which enables accurate detailed definition regions. We achieved good results page-level: 18.71 % for Character Error Rate (CER), 39.7 Word (WER), 14.11 Layout Ordering (LOER), 66.67 mean Average Precision (mAP).