Machine Learning model for Numeral Recognition
Harsha Halesh Patel, Hindu Shree C T, Jayanth S, Keerti D Kulkarni, Pushpa Mala S · 2021 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT) · 2021
Numeral recognition from images of documents has drawn much more attention due to its vast applications involving processing bank cheque leaf, automatic number-plate recognition, identification of ID cards and zip codes. In a computer vision system, numeral recognition is a complex task since the numbers are not written or scripted accurately, as they differ in shape or size, due to which, feature extraction and segmentation of handwritten numerical script is very hard. The aim of the proposed methodology is to build and develop an application such that the model is trained using our own dataset. This ML model can recognize handwritten digits (0 to 9) either through camera or predict the digits drawn on the screen in real time. In this proposed work, optimized python codes are scripted using technologies such as Machine learning, python and its libraries like Scikit-learn, Open CV as well as SVM to develop a training and testing model to produce optimum results.