Driver Fatigue Detection Based on Face Mesh Features Using Deep Learning

The number vehicles and road users increases every year, this also has the potential to increase the risk of traffic accidents. Fatigue is the most dominant cause of accidents compared to several other factors. This research focuses on detecting driver fatigue. We use the mediapipe face mesh model to extract the key points on the face, next is to utilize the deep learning model, Long Short Term Memory (LSTM), which has been trained previously and implemented into mediapipe to detect driver fatigue. The data that is trained is the data point movement of facial features, so that the system can not only process one frame but several frames. The data given by the camera will be processed using the LSTM model's ability to detect long-term information, dynamically process data, and handle picture data using mediapipe in order to achieve low computational and high accuracy. The LSTM model has better accuracy in predicting facial features than the conventional random forest model.