Driver Drowsiness-Detection System based on Transfer Learning and MC-KCF

The driver fatigue is an important issue in many vehicle accidents. According to a study by Transport Research Wing, drowsy driving causes more than 1 lakh crashes, 71 thousand injuries, and nearly 2 thousanddeaths each year in India.Earlier approach methods are based on AlexNet, GoogleNet, MobileNet and Dricare technique (SqueezNet) and uses video images to identify the driver's state of fatigue, such as yawning, blinking, and the length of eye closure, without fitting their bodies with gadgets. The main challenges faced in implementing earlier approach is image classification and increase in percentage error rate when a greater number of layers are added.The proposed method is based on the issues in connection with the driver drowsiness detection and alert device. The proposed algorithm uses the features of deep convolutional neural network-like RESNET (Residual neural networks) and MC-KCF (Multi Convolution neural network-KCF). These techniques are used to detect driver drowsiness by measuring through eye closure and eye rotation. The proposed method helps in better image classification and produces low percentage error rate.