Drowsy Driver Detection System

A system to detect drowsiness using extracted HOG facial landmarks. Developed as a part of CSE 4510: Software Development.

Team Members

• Mashrur Mahmud Morshed(@ID56)
• Hasan Tanvir Iqbal(@TanvirHundredOne)
• Mazharul Islam Rishad(@Aporamithos)

Features

• Detects drowsiness from a real-time video stream
• Calculates number of blinks of driver and shows on output display
• Shows runtime information on display
• Generates an excel file(.csv format) containing metadata:
  • Eye position per three seconds
  • Blinks per ten seconds
  • Drowsiness event timestamp
  • System start and end timestamp
  • Effective frame rate

Current Status

Implementation done on a desktop scale. Will implement in raspberry pi for portability.

Frameworks/Libraries Used

• Python
• OpenCV
• DLIB
• Anaconda distribution

Notes

After installing dependencies, download the pre-trained landmark predictor from the dlib github repository, and keep in the same directory.

Credits

-Davis King, creator of the dlib library.
-The Pyimagesearch BLog, for it's neat drowsiness tutorials.