Drowsy driving is one of the leading causes of serious traffic accidents. Detecting and warning drivers of drowsiness can significantly reduce the risk of accidents, thereby contributing to protecting human life and property.In this article, we introduce a model of Convolutional Neural Networks (CNN) to detect the driver’s drowsiness. A video camera is used to monitor [9] the face, the eye state, the mouth state, and the driver’s head pose. The system will issue a warning when the driver manifests drowsiness. We present a new approach in determining one’s eye, mouth, and head pose in different directions combines with using the support of the OpenCV library to locate the driver’s face with the camera. We combination of eyes status, mouth status, and head pose can increase the accuracy and reliability of the system. The overall accuracy when testing the system is over 97%, which has proved the feasibility of the system. In particular, the accuracy of face recognition is 99% with the Caffe model and the support of the OpenCV. Regarding the recognition of the state of the head pose after testing for accuracy is above 97%. On the recognition of the state of the eyes and mouth are very high. Specifically, the accuracy of the eye is over 99.7% and the accuracy of the mouth is 99.8%.