This project integrates TinyML, embedded AI, and computer vision technologies to provide advanced solutions for real-time image processing and analysis on embedded systems. It features AI-based facial recognition, optimization through a Kalman filter, and IoT communication with an ESP32 microcontroller. Additionally, the system includes comprehensive data analysis and visualization capabilities using Streamlit. Designed for industrial applications, this system enhances operational efficiency and decision-making through real-time monitoring, on-device inference, and optimized communication.

This project consists of several interconnected components:

Ensure you have the following installed:

Python 3.x

Flask

OpenCV

Dlib

Keras

NumPy

Pandas

Plotly

Streamlit

Scikit Learn

Tensorflow

Arduino IDE or ESP-IDF for ESP32 programming

Clone the repository: git clone https://github.com/Fouad-Smaoui/Real-Time-Industrial-AI-Monitoring-System.git cd Real-Time-Industrial-AI-Monitoring-System

Install Python dependencies: pip install -r requirements.txt

Install Arduino IDE or ESP-IDF.

Connect your ESP32 to your computer.

Open main.cpp in the Arduino IDE.

Update the WiFi credentials and MQTT broker address.

Upload the code to the ESP32.

Run the Flask application: python app.py

Run the Jupyter notebook: jupyter notebook data_analysis.ipynb

Run the Streamlit application: streamlit run streamlit_app.py

This notebook demonstrates comprehensive data analysis on industrial sensor data using NumPy for numerical operations, Pandas for data manipulation, and Plotly for interactive visualizations. It includes exploratory data analysis (EDA), statistical experiments, and hypothesis testing to derive insights from sensor readings.

The Flask app captures video, detects faces against known identity, recognizes facial features using a pre-trained Keras model and enhances accuracy through Scikit-Learn's PCA module. The results are displayed in real-time.

The Kalman filter optimizes face tracking by predicting and smoothing the trajectory of detected faces, improving tracking accuracy and robustness under motion.

The ESP32 code configures WiFi, MQTT communication, and runs a FreeRTOS task to handle MQTT operations.

This project includes a TinyML component designed to enable efficient model inference on low-power devices such as the ESP32. By utilizing TensorFlow Lite, we can deploy machine learning models that run directly on the hardware, minimizing latency and reducing the need for cloud-based processing.

Model Training and Conversion: A simple neural network is trained on the MNIST dataset for digit recognition. The trained model is converted to TensorFlow Lite format for compatibility with embedded devices.

The Streamlit app provides an interactive interface for data analysis and visualization using Plotly for dynamic and engaging plots.

This system is designed for industrial environments where real-time monitoring, data analysis, and optimized communication are essential. Potential applications include:

Smart Surveillance: Real-time facial recognition and tracking for enhanced security.

Industrial Automation: Monitoring and controlling industrial processes through IoT and AI.

Predictive Maintenance: Using data analysis to predict equipment failures and optimize maintenance schedules.

Quality Control: Automated inspection and analysis of products using computer vision.

Edge Computing with TinyML: Enabling on-device inference using TensorFlow Lite, allowing for low-latency decision-making in environments with limited connectivity, such as remote industrial sites.

By integrating these technologies, the system provides a robust solution for enhancing operational efficiency and decision-making in industrial settings.

contributions are welcomed from the community. Please follow these steps to contribute:

Fork the repository.

Create a new branch (git checkout -b feature/your-feature).

Commit your changes (git commit -am 'Add your feature').

Push to the branch (git push origin feature/your-feature).

Create a new Pull Request.

This project is licensed under the MIT License. See the LICENSE file for details.