Journal
Scientific and Technical Journal of Information Technologies, Mechanics and Optics
UDK004.932.75
Issue:3 (149)
Download PDF0 Kbyte
Joint recognition of text and layout in historical Russian documents
Annotation
In this paper, we evaluated the Document Attention Network (DAN), the 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 whole documents from any size as an input and output the text as well as logical layout tokens. For comparison purposes, we conduct our experiments on Digital Peter dataset as it has been recognized at line-level. Dataset consists of documents of Peter the Great manuscripts; ground truths are represented according to a sophisticated XML schema which enables an accurate detailed definition of layout and text regions. We achieved good results at page-level: 18.71 % for Character Error Rate (CER), 39.7 % for Word Error Rate (WER), 14.11 % For Layout Ordering Error Rate (LOER), and 66.67 % for mean Average Precision (mAP).
Keywords
Постоянный URL
Articles in current issue
- Analysis of frequency-robust multivariable dynamical systems
- Fractal micro- and nanodendrites of silver, copper and their compounds for photocatalytic water splitting
- Mathematical modelling of tri-layer dielectric OTFT based on pentacene semiconductor for enhancing the electrical characteristics
- Researching carbon dioxide hydrates in thin films via FTIR spectroscopyat temperatures of 11–180 K
- Method for increasing the information value of video data based on the removal of redundant frames and entropy estimation
- Attacker group detection method based on HTTP payload analysis
- Facial keypoints detection using capsule neural networks
- Review of national and international standards for categorizing of critical information infrastructure objects
- Criterion of the network infrastructure security
- A novel approach to feature collection for anomaly detection in Kubernetes environment and agent for metrics collection from Kubernetes nodes
- Time parameters linear approximation method in elastic systems
- Role discovery in node-attributed public transportation networks: the study of Saint Petersburg city open data
- Exploring the possibility of predicting users’ career guidance preferences based on analysis of community topics and the gender in the online social network users’ profiles
- Blindness detection in diabetic retinopathy using Bayesian variant-based connected component algorithm in Keras and TensorFlow
- Intelligent clinical decision support for small patient datasets
- Assessment of the readiness of a computer system for timely servicing of requests when combined with information recovery of memory after failures
- Buckling analysis of an orthotropic cylindrical shell structure in the ANSYS Mechanical APDL software package
- Justification of the choice of mobile broadband access technology for building radio communication networks of railway transport
- Comparative performance analysis of DVR & DSTATCOM for distributed generation with gravitational search algorithm
- Estimation of the moments of a quantized random variable
- Experimental method for estimating the dynamic error of devices and sensors under their operating conditions
- Method of type-C liquified natural gas tank modeling based on volume optimization for future “milk-run” exploitation
- Optical properties of borate family nonlinear crystals and their application in sources of intense terahertz radiation
- A model of a refractive fiber optic sensor sensing element based on MMF-SMF-MMF structure using surface plasmon resonance