Development of An Application Transforming Handwriting into Digital Form using CNN

Josulin, Claudio
Kurniawati, Yulia Ery

Journal of Applied Research In Computer Science and Information Systems
Vol. 1 No. 2 (2023): Desember 2023

This study aims to develop an application to recognize and predict handwriting using a Convolutional Neural Network (CNN) with ResNet50 architecture. The software development life cycle (SDLC) is an incremental model with two increments. The first increment is used to build the model, and the second increment is used to build the user interface. The data used in this study is handwritten images of Latin uppercase, Latin lowercase, and Arabian numerals with 62 classes. The training data used English Handwritten Characters by Dhruvil Dave from Kaggle Dataset. Data was trained and validated using k-fold cross-validation with tenfold and ten epochs for each fold. The model has an accuracy, precision, recall, and f1-score of 66.33%, 73.4%, 66.2%, and 66%, respectively. The functional application can work as expected based on the black box testing. The developed application can predict handwriting with up to 50% accuracy.