A Lightweight Driver Drowsiness Detection System Using 3DCNN With LSTM

Today, fatalities, physical injuries, and significant economic losses occur due to car accidents. Among the leading causes of accidents is drowsiness behind wheel, which can affect any driver. Drowsiness sleepiness often have associated indicators that researchers use identify promptly warn drowsy drivers avoid potential This paper proposes a spatiotemporal model for monitoring visual from videos. depends on integrating 3D convolutional neural network (3D-CNN) long short-term memory (LSTM). The 3DCNN-LSTM analyze sequences by applying 3D-CNN extract features within adjacent frames. learned are then used as input LSTM component modeling high-level temporal features. In addition, we investigate how training proposed be affected changing position batch normalization (BN) layers in units. BN layer examined two different placement settings: before non-linear activation function after function. study was conducted publicly available datasets named 3MDAD YawDD. mainly composed synchronized recorded frontal side views drivers. We show increases convergence speed reduces overfitting one dataset but not other. As result, achieves test detection accuracy 96%, 93%, 90% YawDD, Side-3MDAD, Front-3MDAD, respectively.