Face Mask Detection
System
Yuvraj PassiUsing AI and ML
Shobhit Jain
Shubhang Singh
Agenda
1 2 3 4 5
Introduction Methodology Model Model Results
Architecture Training
2 Presentation title 20XX
Introduction
The COVID-19 pandemic fundamentally changed the way we interacted in
public spaces, making face masks a critical component of public health.
Enforcing the usage of masks, especially in crowded areas, is essential to
prevent any virus transmission.
Manual monitoring, however, is time-consuming and inefficient.
This project addresses the challenge by introducing an automated,
AI-powered face mask detection system that can monitor
compliance in real-time.
3
Related Work
• Several studies and projects have explored the use of computer vision for face
detection and classification. Most notably, advancements in deep learning,
particularly using CNNs, have made it possible to achieve high accuracy in image
classification tasks.
• This project focuses on a custom-trained CNN model optimized for face mask detection.
Unlike previous solutions, this project integrates real-time video stream analysis, making it more
suitable for practical, real-world applications like surveillance in public places.
4 Presentation title 20XX
Methodology
• Data Collection and Preprocessing:
Dataset:
The dataset used for training the model consists of images categorized into two classes:
"With Mask" and "Without Mask".
Data Augmentation:
To enhance model performance and prevent overfitting, data augmentation techniques such as
rotation, flipping, and scaling were applied.
Image Preprocessing:
Images are resized to 128x128 pixels, converted to RGB, and normalized for optimal model
training.
Face Mask Detection
5 20XX
Using AI and ML
A Few Snippets
• Dataset:
6 Presentation title 20XX
Confusion Matrix
7 Presentation title 20XX
Model Accuracy
8 Presentation title 20XX
Model Architecture
• Convolutional Neural Network (CNN): The model is a custom-built CNN designed to detect face masks in real-
time. The architecture includes:
• Input Layer: Accepts images of size 128x128x3.
• Convolutional Layers: Multiple layers with ReLU activation to extract features
• Pooling Layers: MaxPooling layers to reduce spatial dimensions.
• Dense Layers: Fully connected layers leading to a binary classification output.
• Output Layer: Sigmoid activation function for binary classification ("With Mask" or "Without Mask").
Face Mask Detection
9 Using AI and ML 20XX
Model Training and Evaluation
• Training:
The model was trained using the Adam optimizer with a learning rate of 0.001, and binary cross-entropy
as the loss function.
• Evaluation Metrics:
Accuracy, precision, recall, and F1-score were used to evaluate the model's performance on the test set.
• Real-Time Detection:
OpenCV was utilized for capturing live video feed and detecting faces using the Haar Cascade Classifier.
The trained model was then applied to classify detected faces.
Face Mask Detection
10 20XX
Using AI and ML
Live Video Stream Integration
• Webcam Integration: The system uses OpenCV to access the webcam for real-time
mask detection.
• User Interface: The detected faces are highlighted with bounding boxes, and the
classification result ("With Mask" or "Without Mask") is displayed on the video feed.
Face Mask Detection
11 Using AI and ML 20XX
Experiment Results
Training Accuracy: Achieved 98% accuracy on the training dataset.
Validation Accuracy: Maintained an accuracy of 96% on the validation dataset .
Real-Time Detection Accuracy: Successfully classified mask usage with over 95% accuracy in real-time video feeds. The
system was tested under various lighting conditions and angles, demonstrating robust performance in detecting masks.
• Sample Output
• The system draws a bounding box around the detected face and displays a label indicating "With Mask" or "Without
Mask".
The prediction confidence score is also displayed for each detection.
Face Mask Detection
Using AI and ML
12 20XX
Conclusion
The Face Mask Detection System developed in this
project provides a reliable and efficient solution for
monitoring mask compliance in real-time. The system's
high accuracy and responsiveness make it suitable for
deployment in public spaces to ensure health safety. By
leveraging deep learning techniques and real-time video
analysis, this project offers a scalable approach to
automated surveillance
13
Any
Questions?
