Automated Mark Entry in Educational Institutions Using Multiple CNNs: A Case Study on Recognizing Handwritten Fractional Marks
Don Sabu, Aneeta Jose, J. R. Philip, P. Pranav, Renjith Thomas, Jabin Mathew · 2024
This study presents a methodology for the recognition of single-digit with fractional handwritten numbers utilizing multiple specialized Convolutional Neural Network (CNN) models. In contrast to conventional approaches that uses complex segmentation, this method enhances digit recognition by eliminating the requirement for complex preprocessing. The system was designed to automate post-evaluation documentation in educational sectors, addressing the inaccuracies and inefficiencies that come with manual data entry. The experimental result shows an accuracy of 96.7% for fractional numbers and 96.3% for single digits. The modular architecture ensures adaptability and scalability, rendering it suitable for practical implementation in academic environments. This system can be further extended for applications in domains beyond education, it has the potential to be developed to assist students in identifying areas that require more focus.