Optimizing Arabic Handwriting Recognition Using Deep Learning and the Aquila Optimizer

Arabic handwritten characters are crucial in various applications, necessitating highly effective recognition systems. This research presents a novel approach to Arabic Handwritten Recognition (AHR) by leveraging non-segmented techniques. The study utilizes the Aquila Optimizer to enhance the architectures of ConvNeXt and NfNet-F5 deep convolutional neural networks (CNNs), optimizing feature extraction and classification processes. Using the HACDB database, we demonstrate the effectiveness of our method, highlighting significant improvements over existing approaches. Notably, the optimized ConvNeXt architecture achieved an impressive accuracy rate of 98.96%, underscoring the robustness of our approach.