Automated handwritten character recognition (AHCR) is one of the fascinating areas of artificial intelligence and pattern recognition, gaining popularity due to its broad array of applications in medical, finance, biometric authentication, invoicing imaging, computerized number identification, and handwriting recognition. The documents are digitized and converted into editable formats as a result of this recognition procedure. Although the AHCR has a long history, established methods, and configurations, it also presents several advantages and disadvantages. The diverse backdrops, differences in handwriting, font sizes, angles, low contrast, and multi-language text make character recognition in natural images a time-consuming and challenging. Deep-learning algorithms, specifically fully convolutional neural networks (FCNs), have emerged as a viable option for HCR. This paper employs deep learning methods based on fully convolutional networks (FCNs) to develop a novel solution for AHCR. The model was validated using independent dataset testing and 10-fold cross-validation, which yielded accuracy rates of 96.6 percent accuracy, 96.1 percent precession, 97.6 percent recall, and 97 percent F1-Measure, respectively.
Improved Fully Convolutional Neural Networks for Automated Handwritten Character Recognition
2022-12-01
1082596 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Handwritten Character Recognition by Extended Loop Neural Networks
British Library Conference Proceedings | 1994
|Fuzzy Logic in Handwritten Character Recognition
British Library Online Contents | 2003
|Automated Vehicle Recognition with Deep Convolutional Neural Networks
Transportation Research Record | 2017
|Handwritten Hangul Character Recognition with Hierarchical Stochastic Character Representation
British Library Conference Proceedings | 2003
|A New Type of Recurrent Neural Network for Handwritten Character Recognition
British Library Conference Proceedings | 1995
|