Facial Recognition Application – Face Mask Detection with Deep Learning
Project Overview
The Facial Recognition Application is a deep learning-based system designed to detect face masks in real-time. It classifies images into three categories:
-
Wearing Mask Properly
-
Not Wearing Mask
-
Wearing Mask Incorrectly
This application was developed to support public health monitoring, especially during pandemic conditions, by ensuring compliance with mask-wearing guidelines in crowded or indoor environments.
My Role & Key Tasks
As a Full-stack Developer, I independently built the entire system, covering everything from data handling to model training and evaluation. Here’s what I worked on:
-
Data preprocessing: Loaded and augmented image data using ImageDataGenerator, normalized pixel values, and organized dataset into 3 mask categories.
-
Model training: Fine-tuned a pre-trained MobileNetV2 model for the classification task using TensorFlow/Keras.
-
Evaluation: Used confusion matrix and classification reports (via Seaborn and Scikit-learn) to evaluate model accuracy, precision, recall, and F1-score.
-
Model deployment: Exported the trained model (.h5) for future deployment in real-time systems using OpenCV or web frameworks.
Key Features & Highlights
-
Three-Class Mask Detection: Goes beyond binary detection by identifying incorrectly worn masks.
-
Transfer Learning with MobileNetV2: Efficient and lightweight model, ideal for real-time applications.
-
Visual Evaluation Tools: Integrated Seaborn and Matplotlib to generate clear performance metrics and insights.
-
Deployment-ready: Model saved and prepared for integration with camera systems or Flask/React-based web apps.
Results & Accuracy
The model achieved strong accuracy on the test set, with confusion matrix visualizations clearly distinguishing between the three mask-wearing states. Misclassification rates were minimized through image augmentation and class balancing during training.
Conclusion
This facial recognition project represents a meaningful application of AI in healthcare and public safety. It reflects my ability to take a machine learning project from start to finish — combining data science, computer vision, and software development skills into a deployable solution.