Drowsiness Detection for Driving

This project is aimed at detecting driver drowsiness by analyzing eye states (open/closed) and yawns using deep learning techniques. The model is based on YOLOv5 and performs both image classification and object detection to recognize drowsiness signs in real time.

Project Overview

Drowsy driving is a significant factor in road accidents worldwide. Early detection of drowsiness can save lives by alerting the driver and taking appropriate preventive measures. This system identifies drowsiness by monitoring:

1. Eye State: Whether the driver’s eyes are open or closed.
2. Yawns: Detecting when the driver is yawning, which is a strong indicator of tiredness.

Dataset

This project utilizes two key datasets:

• Eye Detection Dataset: Used for detecting eyes in different states (open or closed) with YOLOv5.
• Eye and Yawn Classification Dataset: Used for classifying eye states and detecting yawns for drowsiness analysis.

How the Project Works

1. Eye State Classification:
   • The model classifies images as either "eyes open" or "eyes closed" to determine drowsiness.
   • This classification is crucial as prolonged eye closure is a common sign of drowsiness.
   • Two approaches were used for detecting eyes:
     • Eye Detection Dataset: A bounding box is drawn around the eyes using a specialized eye detection dataset.
     • Face Detection Dataset: A bounding box is drawn around the entire face, and the eyes are detected within that region.
2. Yawn Detection:
   • The YOLOv5 model is used to detect yawns, which often occur when a person is drowsy.
   • The model continuously monitors the driver's face for signs of yawning.
3. Drowsiness Decision Logic:
   • The system integrates the results from both eye state classification and yawn detection to make a final drowsiness prediction.
   • If the eyes are closed for a specified period or if yawns are detected frequently, the system triggers an alert.

YOLOv5 for Drowsiness Detection

YOLOv5 (You Only Look Once) is a real-time object detection system. In this project, YOLOv5 is used for:

• Eye Detection: Identifying the position and state of the eyes (open or closed).
• Yawn Detection: Detecting when the driver is yawning using bounding boxes around the mouth region. The model is trained to be lightweight and efficient, making it suitable for real-time deployment in vehicle systems.

Example Outputs

Here are some sample outputs showing the detection of drowsiness indicators:

using Object detection for eyes and Classification for yawns

using image classification with face detection bounding box

using image classification with eye detection bounding box

Conclusion

This project highlights the application of deep learning for enhancing road safety. By integrating eye state and yawn detection, the system provides a robust method for detecting driver drowsiness and issuing timely alerts. Future improvements could involve adding more features like head pose estimation and facial landmark detection for even greater accuracy.

References

• YOLOv5: Ultralytics YOLOv5
• Dataset 1: Eye Detection
• Dataset 2: Eye and Yawn Classification