A driver fatigue detection algorithm which is based on the states of eyes and mouth of drivers is proposed to detect fatigue state. A multitask cascaded convolutional networks（ MTCNN） is applied for face detection and feature point positioning. After positioning

another convolution neural network（ CNN） is used for eye and mouth regions state recognition. By calculating the percentage of eyelid closure（ PERCLOS） parameter and the frequency of mouth movements

determine fatigue state of driver. The experimental results show that the proposed method has high detecting accuracy and can quickly detect fatigue state of driver and meet requirements of real-time.