Skip to content

README.md

Latest commit

 

History

History
74 lines (48 loc) · 2.26 KB

README.md

File metadata and controls

74 lines (48 loc) · 2.26 KB

Face Detection and Recognition with MTCNN and OpenCV

This Python script provides a full pipeline for face detection and recognition using the MTCNN detector and OpenCV's LBPH (Local Binary Patterns Histograms) face recognizer. It captures photos from a webcam, trains a recognizer on the captured faces, and detects faces in real time, logging detections to a SQLite database.

Features

  • Face detection using MTCNN (Multi-task Cascaded Convolutional Networks)
  • Face recognition using OpenCV's LBPHFaceRecognizer
  • Real-time webcam-based detection and recognition
  • SQLite database logging of detected faces and timestamps
  • Simple user interaction via keyboard commands

Requirements

  • Python 3.x
  • OpenCV (opencv-contrib-python for face recognition module)
  • MTCNN
  • NumPy
  • SQLite3 (standard with Python)

Installation

Install the required packages using pip:

pip install opencv-contrib-python mtcnn numpy

How It Works

1. Capturing Photos

Run the script and enter your name when prompted. Press:

  • s to start capturing your face images.
  • e to stop capturing.

The images will be saved in a folder dataset/{your_name}.

2. Training the Recognizer

Once photo capturing is complete, the script automatically:

  • Loads all grayscale face images from the dataset folder.
  • Labels them numerically.
  • Trains an LBPH face recognizer on this data.
  • Saves the trained model as face_model.yml.

3. Detecting Faces

After training, the script activates webcam-based detection and recognition:

  • Faces are detected in real time.
  • Each detected face is converted to grayscale and recognized using the trained model.
  • If confidence is high (less than 70), the name and timestamp are logged into face_detection.db.
  • The detected face is annotated in the webcam feed with a green rectangle and name.

Press q to quit face detection.

Output Files

  • dataset/: Folder containing captured face images.
  • face_model.yml: Trained LBPH face recognition model.
  • face_detection.db: SQLite database logging detections.

Notes

  • Make sure your lighting is good while capturing images.
  • Capture multiple images (20–30) for better training accuracy.
  • MTCNN might detect false positives in low-light or noisy conditions.

License

MIT License