Arabic handwritten recognition of Tunisian postal addresses using deep learning

Most of the postal systems in Arabic countries like Tunisia are still manually operated for mail sorting and processing. In fact, the processing of Handwritten Arabic Script remains a particularly distinctive problem due to people is varying writing styles. Therefore, we intended to propose an effective and efficient Offline Arabic Handwritten Recognition System of Tunisian Postal Address used a deep learning method (Faster R-CNN). Experimental evaluations demonstrate that the approach is competent and able to accurately detect and classify Tunisian postal code with 84% and either the stamp block with accuracy of 97 %, the Receiver Name with 96% and the Receiver Address with 94%. p. 75-81