Face Mask Detection

Real-time mask detection using VGG16 and HaarCascade

This project demonstrates real-time face mask detection using a transfer learning approach with VGG16 for classification and HaarCascade for face detection.

A notebook version is available below for reproducibility:

Real-Time Detection

OpenCV’s HaarCascade detects faces in the video stream. Detected faces are passed to the trained classifier, and predictions (mask / no mask) are displayed on the live feed.

Key Takeaways

  • Transfer learning with VGG16 achieves reliable mask detection with limited data.
  • HaarCascade ensures lightweight face detection in real time.
  • The system demonstrates practical AI for public health and safety.

Future Work

  • Experiment with alternative pretrained models (e.g., ResNet50, MobileNet).
  • Improve robustness on diverse datasets with occlusion and low light.
  • Deploy on edge devices for real-world use cases.

Resources