Optimizing Arabic Handwriting Recognition: A Deep Learning Approach with Simulated Annealing

Arabic handwritten characters play a crucial role in various applications, necessitating effective recognition systems. This research introduces a novel approach to Arabic Handwritten Recognition (AHR) using non-segmented techniques. Leveraging the Simulated Annealing, we optimized the architectures of EfficientNet B-7 and NfNet-F5 Deep Convolutional Neural Networks (CNNs) to extract features and perform classification. Our investigation focuses on the HACDB database, showcasing the effectiveness of our methodology. Comparative analysis against existing approaches highlights the robustness and efficacy of our proposed method. Remarkably, the optimized NfNet-F5 architecture achieves a remarkable success rate of 97.73%, underscoring the effectiveness of our approach.