The goal of this project is to recognize individual handwritten Arabic letters from images by using convolutional neural networks.
The project uses data to build a convolutional neural network model. The model was trained for 50 epochs. Also, the project uses Adam optimizer. The performance of the model was evaluated based on the accuracy and loss rate of test set.
Hijja is a publicly available dataset of Arabic handwritten alphabets. It consists of 29 characters which are 28 Arabic letters in addition to “Hamza”