Multi-Output CNN for Handwritten Base and Exponent Recognition (Presented at ICSET-2025)

This paper introduces a novel approach for recog-nizing handwritten mathematical expressions, with a particularfocus on predicting both base and exponent values from images.The research utilizes a simplified yet effective method based on amulti-output Convolutional Neural Network (CNN) to accuratelypredict these components. The model is trained on a datasetof 50,000 handwritten images containing exponent expressions,which incorporate random variations in font size and position toclosely mimic real-world conditions. These variations are essentialfor ensuring the model’s robustness and generalization capabili-ties. The proposed CNN model demonstrates strong performance,achieving high accuracy in predicting the base and exponentvalues while maintaining efficient training time. The experimentalresults underscore the model’s effectiveness in handling diverseand challenging input images, making it a valuable tool forhandwritten mathematical expression recognition. This researchcontributes significantly to the field by offering a practical andefficient solution that strikes a balance between model simplicityand performance, making it suitable for real-world applicationswhere computational resources may be limited.