A robust Drowsiness Detection System using two approaches:
- Deep Learning with Transfer Learning (InceptionV3) 🚀
- MediaPipe for Eye Aspect Ratio (EAR) and Head Pose Estimation 📈
This project aims to enhance road safety by detecting driver drowsiness in real-time using computer vision and deep learning techniques. The system alerts drivers when drowsiness is detected to prevent accidents.
-
Real-Time Detection 📹
- Uses webcam feed for real-time monitoring to identify drowsiness.
-
Deep Learning Approach 🧠
- Robust and Flexible: The transfer learning model with InceptionV3 is trained to detect drowsiness under various conditions. It can accurately classify eye states (open/closed) even with head movement, moderate lighting changes, and different face orientations.
- Suitable for applications where precision and adaptability are crucial.
-
MediaPipe Approach 🧩
- Lightweight and Efficient: This method calculates the Eye Aspect Ratio (EAR) and monitors head pose without the need for neural network training.
- Limitations: It works well when the neck is not excessively tilted down, making it ideal for scenarios with minimal head movement.
- Highly efficient, suitable for resource-constrained devices.
-
Custom Alerts 🔊
- Plays an alarm sound if drowsiness is detected for an extended duration.
-
Interactive Visualization 🖼
- Displays live status, EAR values, and predictions with overlays in real-time.
- Achieved high accuracy on test data with fine-tuned InceptionV3.
- Robust to variations in lighting, face orientation, and head movements.
- Ideal for systems requiring precision across diverse environments.
- Lightweight and efficient for devices with limited resources.
- Effective when head movements are minimal and neck tilt is not excessive.
- Suitable for real-time, resource-constrained applications.
For queries, feel free to contact me at: 📧 [email protected]