A Deep Learning Approach to Recognize Telugu Handwritten Numerals
Kranthi Kiran, G. Sai Prakash, V S Monish Ram, Vamshi Krishna Munipalle · 2022 IEEE 3rd Global Conference for Advancement in Technology (GCAT) · 2022
Numerous languages are now recognized by numerical recognition systems due to the rise in international commerce and correspondence, especially in multilingual countries like India where many languages are spoken simultaneously. Every language has its own set of characters and numerals, so every language requires a model to recognize the characters and numerals written in the language. Many models have been proposed in the field of handwritten numeral recognition. But the research done in Indian languages is very limited because of the unavailability of datasets. Telugu is one of the Indian languages with one or two models proposed up to now for handwritten numeral recognition, but the datasets used for the development of those models are not robust. We created a dataset of Telugu numerals with 2250 samples and developed a CNN model that can recognize the handwritten Telugu numerals with an accuracy of around 97%. We also tried to recognize the multi-digit Telugu numerals with help of contour detection.