Drowsiness-Detection-and-Head-pose-Estimation

Drowsiness detection and Head pose Estimation in realtime using Laptop's camera. The project uses python, openCV and mediapipe.

Demo

Showcasing the feature https://github.com/KarimIbrahim11/Drowsiness-Detection-and-Head-pose-Estimation/assets/47744559/59f9e0bb-690a-4ca4-b3cb-47d6f7386461

Solution

1- Face Mesh detection using mediapipe 2- The facial landmark points "left": [362, 385, 387, 263, 373, 380] & "right": [33, 160, 158, 133, 153, 144] were chosen to calculate the inter-eyed distance and the EAR. 3- EAR fixed is 0.25 and the time is 2 Seconds to start reporting Drowsiness. 4- For Headpose estimation, I started off by calculating the rotation and translation vectors between 2d and 3d facial landmarks using cv2.solvePnP() function 5- the rotation matrix was then decomposed to find the angles x, y, z 6- Angles solely, were used statically in the code to identify the head orientation (nominal,up,down,right,left) 7- The angles were then used to display the 3 coordinate vectors.

Assumptions and Areas of improvement

• Focal Length, Camera matrix and Distortion Matrix were all assumed and were not calculated.
• Currently running 37 FPS. Room for improvement.

Results duplication

1- create a virtual env, I personaly use conda. 2- conda env create -f environment.yaml 3- conda activate env 4- python main.py