Jawi Character Recognition using CNNs

This work explores the use of Convolutional Neural Networks (CNNs), to recognize Jawi handwritten characters. Unlike traditional machine learning, which uses algorithms to analyze and learn from data to make decisions, deep learning structures algorithms into layers to create an artificial neural network capable of independent learning. The project utilizes a dataset of 35 Jawi characters, which includes 29 characters derived from the Arabic alphabet and 6 additional characters specific to the Malay language. The main objectives of the study were to create a Jawi handwritten character database, develop a new deeplearning algorithm for recognizing these characters, and assess the accuracy of the recognition process. One challenge in this task is the variation in handwriting, as individuals have distinct writing styles. However, the CNNs model showed impressive results, achieving an accuracy of over $90 \%$. The report also compares the performance of CNNs with traditional neural networks, highlighting the superior accuracy of CNNs. In conclusion, the study demonstrates that deep learning, particularly CNNs, is highly effective in recognizing Jawi handwritten characters, even with the challenges of diverse handwriting styles, making it a promising solution for similar tasks in other scripts.