Face Mask Object Detection

This is a live, real-time face mask detection system. It uses a two-stage pipeline to identify whether individuals in images are wearing face masks. The model leverages transfer learning with MobileNet to classify facial images into two categories: With Mask and Without Mask. It was trained on a balanced dataset of 10,992 images using TensorFlow and Keras.

Face detection is handled separately using a Haar Cascade Classifier, allowing the system to locate faces before passing them to the classifier. I’ve included the trained CNN model (classifier_model.h5) and supporting files for running detection and classification tasks (classifierModel.py, detectorModel.py, and haarcascade_frontalface_default.xml).

Performance Metrics

Model accuracy: Above 90%

Testing the Model with 6 New Images

Evaluating the Completed Object Detection Pipeline

Real Time Face Mask Detection

Dataset & Dataset Structure

face-mask-12k-images-dataset

Dataset/
├── Train/
│   ├── WithMask/
│   └── WithoutMask/
└── Test/
    ├── WithMask/
    └── WithoutMask/

Requirements

Python 3.10.16 
keras 2.10.0
matplotlib 3.10.3
numpy 1.23.5
opencv-python 4.11.0.86
pillow 11.2.1
scikit-learn 1.6.1
tensorflow 2.10.0