This project focuses on developing a robust face recognition and greeting model using computer vision tools like OpenCV, DeepFace, and TensorFlow, integrated with hardware components such as Raspberry Pi. The primary objective is to accurately identify individuals from a database and greet them by name in real-time. This system holds significant potential for deployment in various environments including offices, hospitals, and other public spaces for purposes ranging from enhanced personalized interactions to security and attendance monitoring.
- Face Detection: Utilized OpenCV's
haarcascade_frontalface_default.xmlfor human face detection. - Visualization: Employed Matplotlib for live footage visualization.
- Model Training: Implemented K-Nearest Neighbors (KNN) and Convolutional Neural Network (CNN) algorithms for real-time face detection.
- Hardware Setup: Raspberry Pi module managed processing, training, and model deployment, supported by the Pi Camera Module 2 for image capture.
- Model Enhancement: Shifted to training the model offline using the Random Forest algorithm to mitigate lag issues.
- Image Preprocessing: Integrated Pillow library for image preprocessing to accelerate face detection and streamline real-time processing.
- UI Integration: Developed a user-friendly interface using Flask as the backend server.
- Frontend Tools: Leveraged HTML, CSS, Bootstrap, and JavaScript for frontend development to enhance user interaction and display real-time face detection results.
- Advanced Model Integration: Transitioned to DeepFace, achieving approximately 88% precision for a larger database.
- Hardware Challenges: Due to Raspberry Pi's hardware limitations, explored alternatives like Grove Vision AI V2 with ARM Cortex-M55 processor and ARM Ethos-U55 NPU, supported by Xiao ESP32S3 for wireless communication.
- Cost-Effective Solution: Considered Realtek AMB82-Mini 8-in-1 SoC for embedded AI capabilities and efficient high-resolution processing up to 2K at 0.4 TOPs, ensuring cost-effective deployment.
- img folder: Contains processed images of individuals with correct labeling and timestamps. These images are used for training the ML model.
- People folder: Contains images directly uploaded by users for future reference. If users want to delete an image, they can remove all its references within this folder.
- Templates folder: Contains HTML, CSS, and JavaScript files used for creating the user interface.
- Uploads folder: Contains images during the processing time.
- app.py: This is the backend file that hosts the user interface. Launch this file to run the application on
localhost:5000. This can be changed once the model is deployed and has a domain. - final_model_rf.py: Contains the ML model that detects faces in real-time.
- Run the Server: Navigate to the directory where
app.pyis located. In the terminal, execute:python3 app.py
- Formatting: Added code block formatting for the command.
- Clarity: Improved phrasing for better readability and consistency.
- Step-by-Step: Kept the steps clear and actionable.
The project presents a robust face recognition and greeting system developed using advanced computer vision tools such as OpenCV, DeepFace, and TensorFlow, integrated with the Raspberry Pi hardware platform. The primary goal was to achieve accurate real-time identification of individuals from a pre-constructed database and greet them by name. This system has significant potential for deployment in various environments, including offices, hospitals, and public spaces, with applications ranging from personalized interactions to enhanced security and attendance monitoring.