An Enhanced Machine Learning Technique for Text Detection using Keras Sequential model

• Published in: 2023 Second International Conference on Electronics and Renewable Systems (ICEARS) pp. 1147 - 1151
• Main Authors: Deepa, R., Gayathri, S., Chitra, P., Jasmine, J. Jeno, Devi, R. Renuga, Thilagavathy, A.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 02-03-2023
• Subjects: Convolutional Neural Networks (CNN); handwritten text; Machine learning; Neural networks; Optical character recognition; Optical Character Recognition (OCR); Optical losses; Python; Renewable energy sources; Text recognition; Training
• DOI: 10.1109/ICEARS56392.2023.10085174

Optical Character Recognition (OCR) has become the well known accepted automated recognition technology that can be used within a variety of functions that converts documents or pictures to modifiable as well as analyzed information. Because they possess a clearly specified form and size, characters that are in print are simple to identify. But is never the situation with handwritten writing. Because each person's handwriting is unique, OCR has trouble reading the characters. This study provides a two-stage process in this work for the identification and categorization of handwritten digits. Convolutional Neural Networks (CNNs) serve as the basis for the categorization. The recognition of the input number is the initial step in the process. The second step is to determine the language the input number originated from. Python is used to carry out the research work on the MNIST database of handwritten numbers. The simulation's findings demonstrate an efficient performance with a near-perfect identification rate. This study has attained 99.5% training accuracy, 99% testing accuracy, and a training loss of 1.5%.