A Fatigue Detection Model Based on Convolutional Neural Network

The use of vehicles has grown steadily during the past few years. This rise in vehicle usage does not correspond to the degree of driving safety. This is seen in Indonesia's 4% average annual increase in accidents. A drowsy driver was one of the human variables that contributed most to this event. In this study, we address this issue utilizing our CNN-based model that was developed using the National Tsing Hua University Driver Drowsiness (NTHU DD) dataset. The model was trained using approximately 600,000 images extracted from the video dataset. The CNN-based model can obtain a 70% accuracy rate for the validation and a 56% accuracy rate for testing.