KhattTech is a mobile application designed for Arabic Handwritten Recognition (AHR) through a deep learning model, enabling users to input images via camera or upload functionalities. The app includes tools for image editing, providing a user-friendly interface for recognizing, editing, and sharing Arabic text in various formats. This report aims to detail the development, implementation, analysis, inference, and deployment of the AHR model.

First, the KFUPM Handwritten Arabic TexT (KHATT) database goes through analysis, and preprocessing. Then, two experiments are conducted, the model is trained from scratch first. The first one uses only Deep Convolutional Neural Networks (DCNN) on Categorical Cross Entropy (CCE) loss and results in an overfitted model. While the second one utilized DCNN for feature extraction, with Bidirectional Long-Short Term Memory (BLSTM) and Connectionist Temporal Classification (CTC) loss and resulted in good fitting with the data. That is an image-based sequence recognition segmentation-free (Line-level) framework. Finally advanced image processing techniques, including image Filtering, image Transformation, and Line Segmentation, are applied, preparing the data for model inference.

As for the usage, this app is used in many fields for digitizing, documentation, archiving, export text to translating, banking and other fields. For instance, teachers use KhattTech to assist them in reading their students' handwritten papers. Some students struggle to understand lecture notes that they wrote in a hurry, this app will be able to help them analyze the notes written. A general usage of KhattTech is making the image searchable, which assists with information retrieval and is editable, saving time and effort.

Historically, Arabic has faced challenges in digital recognition due to its nature, and to the hegemony of the English language in the industry, resulting in marginalization towards it. KhattTech combats this marginalization by providing a solution that empowers users with extracting Arabic text into various digital applications, thereby challenging the historical marginalization of Arabic.