Real-Time Arabic Letter Classification Using TinyML, KAUST

This repository contains the source code and deployment files for real-time Arabic letter classification using TinyML, specifically optimized for Nicla Vision. The model was trained and deployed using Edge Impulse, achieving a test accuracy of 94.33%.

To deploy the model onto Arduino Nicla Vision you can follow these steps:

What You Need

• Nicla Vision board
• Edge Impulse Exported ZIP File (TinyML_Arabic.zip)
• Arduino IDE or Edge Impulse CLI
• USB-C Cable to connect Nicla Vision

Flashing Using Arduino IDE

1. Install Dependencies:

   • Install Arduino IDE
   • Install Arduino Nicla Vision Board Package
   • Install Edge Impulse Arduino Library

2. Open Arduino IDE

   • Go to Sketch > Include Library > Add .ZIP Library
   • Select TinyML_Arabic.zip to import it.

3. Upload the Model

   • Open File > Examples > Edge Impulse > TinyML_Arabic
   • Click Upload to flash the model onto Nicla Vision.

Running the Model

Once flashed, Nicla Vision will capture handwriting input and classify Arabic letters in real-time. The predictions will appear in the Serial Monitor.
Link for the demo to our project: video

• Contributers:

Abdulrahman Alfrahidi

Nayef Aljhani

Mohamed Eltayeb