Face Recognition Using the KNN Algorithm

This project implements a real-time face recognition system using a camera, the OpenCV library, the face_recognition library, and the K-Nearest Neighbors (KNN) algorithm. The system was developed as an academic project in the field of artificial intelligence.

🔧 Technologies

• Python 3.x
• OpenCV
• face_recognition
• NumPy

📁 Project Structure

projekat-knn/
├── main.py                  # Recording and encoding user faces
├── rec.py                   # Real-time face recognition
├── haarcascade_frontalface_default.xml  # Face detection model
├── data/                    # Folder with encoded .npy files
├── README.md                # Project description
├── requirements.txt         # List of required Python libraries

📸 Features

• Recording and encoding user faces
• Saving feature vectors to a .npy file
• Loading the database of known faces
• Face recognition using a manually implemented KNN algorithm
• Displaying the person's name above the detected face

▶️ Running the Project

1. Install the required libraries:

pip install -r requirements.txt

2. First run loadData.py to capture and save face samples.
3. Then run recSystem.py for face recognition.

📂 Dataset

• .npy files are automatically created for each user and saved in the ./data/ folder.
• Each file contains 20 encoded face samples.

⚠️ Note

• The haarcascade_frontalface_default.xml file must be in the same folder as the scripts.
• The application uses the default webcam (device 0).

🖼️ Face Recognition Examples

Below are images illustrating the application running in real time:

These images show an example where the application detects a user's face via the camera and successfully identifies the person based on previously recorded samples using the KNN algorithm.

📄 Author

Milutin Jovanović
Faculty of Electronic Engineering, University of Niš
2024/2025 – Artificial Intelligence