8 SEM REPORT-for the presentation of the

DROWSINESS DETECTION SYSTEM
ENHANCING ROAD SAFETY WITH REAL-TIME MONITORING
BY ANISH KARKI AND ADESH BASYEL

INTRODUCTION
Importance of drowsiness
detection in preventing
accidents.
System uses cameras to monitor
driver's eye movements and
facial expressions.
Real-time analysis with machine
learning algorithms.
Goals: Enhance road safety and
reduce fatigue-related accidents.

PROBLEM OBJECTIVE
Accuracy Issues: Current
drowsiness detection systems
often suffer from high rates of false
positives and false negatives,
leading to unreliable alerts.
Safety Risks: Drowsy driving
significantly increases the risk of
accidents due to slower reaction
times, impaired judgment, and
reduced vigilance.

OBJECTIVES
 Real-time detection of drowsiness.
 Improve driver safety with timely alerts.
 Reduce fatigue-related accidents.
 Use advanced machine learning algorithms.
 Develop a user-friendly system.

REQUIREMENT ANALYSIS
FUNCTIONAL REQUIREMENTS NON-FUNCTIONAL REQUIREMENTS
 Physiological Monitoring: Monitor eye movements.
 Behavioral Monitoring: Track eye movements, blink
duration, gaze direction, facial expressions, and eyelid
closure.
 Data Processing: Real-time processing and machine
learning integration.
 Alert Mechanisms: Provide auditory alerts.
 Performance: High accuracy, low latency, and reliability
under various conditions.
 Usability: Ease of use, comfort, and customization.
 Scalability: Integration with existing vehicle systems and
extendibility for future updates.
 Security: Data protection and system security.
 Maintainability: Easy maintenance and access to technical
support.

ALGORITHM DETAILS
 Convolutional Neural Networks (CNNs):
 Used for real-time eye closure detection.
 Captures frames of the driver's face using a camera.
 Processes images to determine eye states (open or closed).
 SupportVector Machines (SVMs):
 Suitable for binary classification tasks.
 Analyzes facial landmarks such as eye closure, yawning, and head orientation.
 Detects drowsiness by combining feature extraction techniques.
 YOLOv8 (You Only Look Once version 8):
 State-of-the-art object detection model.
 Used for real-time object detection, image classification, and instance
segmentation.
 Improved speed and accuracy over previous versions.

IMPLEMENTATION TOOLS
 OpenCV:
 Image and video processing.
 Facial recognition.
 YOLOv8:
 Real-time object detection.
 High speed and accuracy.
 Python:
 Versatile programming language.
 Integration with OpenCV and YOLOv8.
 Label img:
 Annotation tool for creating datasets.
 Generates text files with class labels and coordinates.

USER TESTING
Test
No
Test Case Precondition Input Data Expected
Result
Actual
Result
Pass/
Fail
1 Verify image resizing Image input available Sample image Image resized to correct dimensions Image
resized to correct dimensions
Pass
2 Validate normalization Image loaded Sample image Pixel values normalized to
0-1
Pixel values normalized to
0-1
Pass
3 Test model loading Model path available N/A Model loads without errors Model loads without errors Pass
4 Validate output shape Model loaded Sample input Output dimensions match expected sizes Output dimensions match expected sizes Pass
5 Accurate predictions Model loaded Known
input
Expected bounding boxes returned Expected bounding boxes returned Pass
6 Test alert triggering Drowsiness
state determined
Drowsy
state input
Alert triggered correctly Alert triggered correctly Pass

SYSTEM TESTING
Test No Test Case Expected Result Actual Result Pass/Fail`
1 Validate image
processing flow
Input flows through preprocessing,
inference, detection
Input flows through preprocessing, inference,
detection
Pass
2 Check output
consistency
Correct alerts produced for various test
inputs
Correct alerts produced for various test inputs Pass
3 Validate performance
under varying
conditions
System performs accurately under
different conditions
System performs accurately under different
conditions
Pass
4 Measure system
accuracy
Accurate detection and response time
under scenarios
Accurate detection and response time under
scenarios
Pass

RESULT ANALYSIS
 Confusion Matrix:
 Visualize true positives, true negatives, false positives, and false negatives.
 Identify areas for improvement.
 Testing on Diverse Datasets:
 Evaluate performance under varied conditions (lighting, facial features, driving conditions).
 Assess model's generalizability and robustness.
 System Usability:
 Gather user feedback on intrusiveness and convenience.
 Ensure the system is user-friendly and non-intrusive.
 Accuracy:
 Measure overall correctness in detecting drowsy vs. non-drowsy states.
 Aim for high accuracy with low false-positive and false-negative rates.

CONCLUSION
 Summary: The Drowsiness Detection System enhances road safety by using advanced technologies to monitor and
alert drivers about drowsiness, reducing the risk of accidents. This user-friendly system achieves high accuracy through
real-time video processing and machine learning, ensuring timely alerts and promoting overall safety.
 Expected Outcomes:
 Accurate detection with low false-positive and false-negative rates.
 Timely alerts to the driver.

8 SEM REPORT-for the presentation of the

8 SEM REPORT-for the presentation of the

More Related Content

Similar to 8 SEM REPORT-for the presentation of the (20)

Recently uploaded (20)

8 SEM REPORT-for the presentation of the

Editor's Notes