Deteksi Kantuk Pengemudi Menggunakan Deep Learning

Salamet Nur Himawan; Robieth Sohiburroyan; Nur Budi Nugraha

Salamet Nur Himawan Politeknik Negeri Indramayu
Robieth Sohiburroyan Politeknik Negeri Indramayu
Nur Budi Nugraha Politeknik Negeri Indramayu

Keywords: Deep Learning, CNN, FCN, Kantuk

Abstract

One of the causes of driving accidents is driver fatigue. Driver fatigue can be measured by the level of sleepiness of the driver. In this study, a deep learning model was developed to detect sleepiness in drivers. The model will detect images of the eyelids and yawning state of the driver. The driver's level of sleepiness was grouped into not sleepy, sleepy and very sleepy. The developed model consists of a feature extractor in the form of several CNN layers, and a predictor in the form of FCN. CNN is able to extract features in the image so that it can provide better accuracy at the prediction stage. The model successfully learns from data and is able to detect driver sleepiness. The accuracy of the model reaches 0.932%. The failure of the model in detecting is caused by the wrong angle of taking the image to capture the front face. It is hoped that for the next research the number of datasets will be increased by all angles of image taking.