Emotion AI Assistant

A real-time facial emotion recognition system built with Raspberry Pi 4 and Arduino Uno R4 WiFi. The system uses a specialized neural network to detect emotions and synchronizes the results with a remote hardware actuator via MQTT.

System Architecture

The project is divided into two main nodes that communicate over the cloud:

1. The Brain (Raspberry Pi 4):
   • Captures images via Picamera2.
   • Processes images using a MobileNetV2 (TFLite) model.
   • Saves logs to an SQLite database.
   • Publishes results to the HiveMQ MQTT Broker.
2. The Visualizer (Arduino Uno R4 WiFi):
   • Subscribes to the MQTT topic.
   • Triggers physical feedback (RGB LED and Buzzer) based on the received emotion.

Tech Stack

• Language: Python 3.x, C++ (Arduino)
• Frameworks: Flask, Flask-SQLAlchemy
• AI/ML: TensorFlow Lite, NumPy, Pillow
• Communication: Paho-MQTT, PubSubClient
• Hardware: Raspberry Pi 4, Picamera2, Arduino Uno R4 WiFi, RGB LED, Active Buzzer

Clone & Setup

• Clone the repository
• Install the required Python packages (pip install -r requirements.txt)
• Run the application (python main.py)

Arduino Configuration

• Open arduino/emotion_node.ino in the Arduino IDE.

• Install WiFiS3 and PubSubClient libraries.

• Update your WiFi credentials (ssid, password) in the script.

• Flash the code to your Arduino Uno R4 WiFi.