Bangla Handwritten Digit Recognition Using Convolutional Neural Network

Abstract

Handwritten digit recognition is the process of automatically identifying and interpreting handwritten digits, such as those found on a bank check or a handwritten phone number. In Bangladesh, this technology could be used in a variety of ways, including automating the process of reading and interpreting documents, reducing the need for manual data entry, and improving the accuracy and efficiency of data processing tasks. It could also be used to help automate certain business processes, such as invoicing and billing, or to streamline the process of filling out forms. Overall, the use of handwritten digit recognition technology could potentially have a significant impact on productivity and efficiency in Bangladesh. Handwritten digit recognition is a machine learning task that involves identifying and detecting written digits from digital data. In order to improve efficiency and transition to paperless offices, a system for recognizing Bengali handwritten digits is necessary. However, recognizing these digits can be difficult due to variations in shape, size, and writing style. In this report, we propose a neural network approach using TensorFlow's CNN algorithm (Keras Sequential Model API) and the BHAND dataset to accurately and efficiently recognize Bengali handwritten digits. Our model, which consists of six layers, achieved a recognition accuracy of 99.1%.