The document outlines the development of a driver drowsiness detection system that uses image processing techniques to monitor eye states and alert drivers of fatigue. By tracking the duration of eye closure, the system aims to prevent accidents caused by drowsiness, achieving a high accuracy of 99%. The project incorporates a convolutional neural network (CNN) for real-time analysis, ensuring reliable monitoring and enhancing road safety.

1. SMART DRIVER ALERT:
PREDICTIVE FATIGUE
DETECTION TECHNOLOGY
VINNEY B JONES

2. PROJECT SCOPE
Reason
Major studies have suggested that
around 20% of all road accidents are
fatigue related. Drowsy Driving can be
extremely dangerous, a lot of road
accidents are related to the driver falling
asleep while driving and subsequently
losing control of the vehicle. However,
initial signs of fatigue and drowsiness
can be detected before a critical
situation arises. Driver drowsiness
detection is a vehicle safety technology
that helps to prevent accidents caused
by driver getting drowsy.
Aim
The aim of this project is to design and develop a driver
drowsiness detection system that utilizes image processing
techniques to monitor the driver's eyes and detect signs of
fatigue. By analyzing eye states, the system will track the duration
of eye closure, and if the driver's eyes remain closed for more
than 15 seconds, an alert will be triggered through the vehicle's
speaker, waking the driver and preventing potential accidents
caused by drowsiness.

3. PROBLEM IDENTIFICATION
Have slept for fewer
than 7 to 8 hours.
Use drowsiness-inducing
medicine.
Drive late at night or
early in the daytime.
Have trouble falling or
staying asleep at night
on a regular basis.
Drive for lengthy periods
of time on monotonous
highways or rural roads.
Work the night shift,
especially if you're
driving home afterward

4. A WALK THROUGH INTERESTING STATISTICS
BASED ON NATIONAL SLEEP FOUNDATION’S DROWSY DRIVING SURVEY AND REPORTS

5. DRAWBACKS FOR
EXISTING MODEL
ADVANTAGES OF PROPOSED SYSTEM
• Ability to Monitor a Wide Range of Indicators Simultaneously
• High Accuracy in Identifying Drowsiness Indicators
• Non-Intrusive and Passive Monitoring
• Cost Effective
• Limited Accuracy in Real-World
Conditions
• Intrusiveness
• High in computational complexity may
lead wrong classification.
• Dependency on Driver Behavior
• High Cost

6. PROPOSED WORK
• This project is to design and
develop a driver drowsiness
detection system that utilizes
image processing techniques to
monitor the driver's eyes and
detect signs of fatigue. By
analyzing eye states, the system
will track the duration of eye
closure, and if the driver's eyes
remain closed for more than 15
seconds, an alert will be triggered
through the vehicle's speaker,
waking the driver and preventing
potential accidents caused by
drowsiness.

7. APPROACH
In this project, we will be using OpenCV for gathering the images from
webcam and feed them into a Deep Learning model which will classify
whether the person’s eyes are ‘Open’ or ‘Closed’. The approach we will be
using for this project is as follows :
Step 1 Capture Image : Collect image from webcam in real-
time.
Step 2 Face Detection : Detect face and define Region of
Interest (ROI).
Step 3 : Eye Detection : Identify eyes within ROI for further
analysis.
Step 4 : Classify Eye Status : Use a Deep Learning model to
classify eyes as "Open" or "Closed".
Step 5 : Drowsiness Score : Calculate score based on eye status
to assess drowsiness.

8. PREREQUISITES & RESOURCES
Software Requirements
• Python - Recommented Version(3.6)
• OpenCV – face and eye detection.
• TensorFlow – keras uses TensorFlow as backend.
• Keras – to build our classification model.
• Pygame – to play alarm sound.
Hardware Requirements
• Processor : i3 or grater
• RAM : 4GB or greater
• Hard Disk : 50 GB or greater
• Connectivity : LAN or WIFI, Camera

9. DRIVER DROWSINESS DATA
• Source : Public Domain
• Description : The dataset consists of 2900 images which
include both open and closed eyes images. Dataset is divided
into training and testing, which is used in the project for
training and testing respectively.
• Image Reference:
Image1 – Open Eye
Image2 – Closed Eye

10. DESIGN CHOOSES :
CONVOLUTIONAL NEURAL
NETWORK (CNN) DESIGN
CNN Architecture
• Input Layer: 32x32x3 image data.
• Convolutional Layers: Extract hierarchical features.
• Pooling Layers: Reduce spatial dimensions.
• Fully Connected Layers: Classify features.
• Output Layer: Class prediction.
Advantages of CNNs
• Automatic feature extraction
• Reduced number of parameters
• Improved accuracy for image-based tasks
What is CNN?
• A deep learning model used for
image and visual data analysis.
• Combines convolution and
pooling operations to extract
features.

11. MODULES USED
2. Face Detection Module
• The module takes an image (frame) and converts it to grayscale
(gray).
• The Haar Cascade classifier (face.detectMultiScale) is applied to
detect faces in the image.
• For each detected face, a bounding box (rectangle) is drawn
around it on the original image (frame).
• The module outputs the image (frame) with rectangles drawn
around each detected face.
1.Model Training Module
• The goal is to develop an intelligent system that detects
driver fatigue by analyzing eye states and facial features.
• Diverse and representative data, such as images of drivers
with both open and closed eyes, is collected to train the
model for accurate fatigue detection.
• Convolutional Neural Network (CNN), is chosen and trained
on the collected data to learn patterns in eye states and
facial expressions.
• Once trained, the system provides real-time feedback,
detecting signs of drowsiness and alerting the driver to
prevent accidents.

12. MODULES USED
4. Drowsiness Detection Module
• Determines whether the eyes are open or closed.
• A score is maintained based on the duration of eye closure.
• If the score exceeds a threshold (e.g., 15), the alert is triggered.
3. Eye Detection Module
• The module takes an image (frame) and converts it to
grayscale (gray).
• Haar Cascade classifier to detect the left and Right eye region
within the image.
• The detected eye region is extracted, converted to grayscale,
resized, normalized, and reshaped to match the input
requirements of the model.
• The processed eye image is passed through a pre-trained
model to predict whether the eye is open or closed.
• Based on the prediction, the eye is labeled as either Open or
Closed.

13. ARCHITECTURE
DIAGRAM
• Process Flow Diagram

14. MODEL ARCHITECTURE
The model we used is built with Keras using Convolutional Neural
Networks (CNN). A convolutional neural network is a special type of
deep neural network which performs extremely well for image
classification purposes. A CNN basically consists of an input layer, an
output layer and a hidden layer which can have multiple layers. A
convolution operation is performed on these layers using a filter that
performs 2D matrix multiplication on the layer and filter.
The CNN model architecture consists of the following layers:
• Convolutional layer; 32 nodes, kernel size 3
• Convolutional layer; 32 nodes, kernel size 3
• Convolutional layer; 64 nodes, kernel size 3
• Fully connected layer; 128 nodes
• The final layer is also a fully connected layer with 2 nodes. A Relu
activation function is used in all the layers except the output layer
in which we used Softmax.

15. MODEL SUMMARY
• CNN layer Summary

16. MODEL ACCURACY PLOTS

17. OUTPUT SCREENSHOTS
Image 1 - Open
Eyes Detection
Image2 - Closed
Eyes Detection
Image 3 - Sleep
Alert

18. SUMMARY
Goal of the System: The Driver
Drowsiness Detection System aims to
enhance road safety by monitoring
and detecting signs of fatigue in
drivers.
Real-time Analysis: The system
analyses the driver’s eye states (open
or closed) and other facial features in
real time to assess alertness.
Data Collection and Preprocessing: A
labelled dataset containing images of
drivers with open and closed eyes is
collected and preprocessed for
training a deep learning model.
Model Training: A Convolutional
Neural Network (CNN) is typically
used to train the model to recognize
patterns in facial features that
indicate drowsiness.
High Accuracy: After training, the
model can detect signs of drowsiness
with high accuracy, providing
immediate feedback to the driver to
prevent fatigue-related incidents.

19. CONCLUSION
• High Accuracy: The Driver Drowsiness Detection
System achieves 99% accuracy, demonstrating
exceptional performance in real-time fatigue
detection.
• Reliable Monitoring: This high accuracy ensures
consistent and reliable monitoring of the driver’s
alertness, allowing the system to effectively
detect drowsiness.
• Preventing Accidents: The system acts as a vital
tool for preventing drowsiness-related accidents,
by alerting the driver when fatigue is detected.
• Timely Alerts: It provides visual or audible signals
to warn the driver, encouraging them to take
actions such as resting or stopping when
necessary.
• Improving Road Safety: The implementation of
this system significantly contributes to safer
roads, helping reduce traffic-related fatalities
caused by driver fatigue

20. FATIGUE-FREE
FUTURES:
REVOLUTIONIZING
SAFETY ACROSS
SECTORS
Healthcare and Sleep
Disorder Monitoring
Workplace Fatigue
Detection for Industrial
and Manufacturing Jobs
Fatigue Monitoring for
Pilots and Air Traffic
Controllers
Home Monitoring
Systems for Elderly Care
Safety Monitoring in
High-Risk Jobs
(Firefighters, Emergency
Responders)
Smart Vehicle/Transport
Services for Elderly or
Disabled Individuals

21. TIPS TO PREVENT
DROWSY DRIVING
• Fatigue Doesn’t Take a
Break, But You Should!