HANDWRITTEN CHARACTER RECOGNITION USING MACHINE LEARNING AND DEEP LEARNING

Abstract

The process of detecting and converting characters from pictures, documents, and other sources into machine-readable structures for subsequent processing is known as handwritten character recognition (HCR). Recognizing intricately constructed compound handwritten characters properly remains a significant challenge. Convolutional neural networks (CNN) have made significant advances in HCR by learning discriminating properties from massive amounts of raw data. In this study, CNN is used to recognise characters in a test dataset. The main goal of this research is to look into CNN's ability to recognise characters from image collections, as well as the recognition accuracy during training and testing. We tested our CNN implementation against the A-Z Handwritten Data to measure the correctness of handwritten characters. dataset in csv. On 372450 images of alphabets in 28*2, the test result shows accuracy of 97.75%.