Using facial landmarks to detect driver fatigue

Martinez A.; Berrospi F.; Porras V.; Portocarrero M. · Proceedings of the 2022 IEEE 29th International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2022 (2022) · DOI: 10.1109/intercon55795.2022.9870046

Resumen

Using state-of-the-art computer vision techniques, a model was designed to be able to detect fatigue in haul truck drivers in South American mines. The proposed solution uses a DNN-based (Deep Neural Networks) face detection model, caffe, to detect the presence of the driver's face and a facial landmark detection model of MobileNetV2 architecture to identify key features of the face. Afterwards, the eye aspect ratio (EAR) and the mouth aspect ratio (MAR) are calculated, which are then used to calculate the percent eye-closure over a time period (PERCLOS), blink rates and duration, yawns, distraction, and other indicators. The results successfully detect fatigue symptoms over 72% of the times on haul truck drivers in Peruvian mines.