Driver Fatigue Detection Based On Eye State Recognition Using Convolutional Neural Network

Fatigue driving is suspected to be a primary cause of vehicle related crashes. Driver fatigue resulting from lack of rest is a significant factor in the expanding number of the mishaps on roads. A warning system can possibly reduce the accidents related to drivers drowsiness. By putting the camera inside the vehicle, the system can read the expressions of the driver and search for the eye-developments which demonstrate that the driver is never again in condition to drive. In such a case, a warning sign ought to be issued. Our face contains a great deal of supportive data; we can utilize the condition of eyes to discover the tiredness. In this paper, a methodology has been proposed for the eye state recognition dependent on convolutional neural network (CNN), which eventually calculates the percentage of eyelid closure (PERCLOS), and blink frequency to figure out the level of the drowsiness. A real-time monitoring unit is additionally created to alert the closest control unit about the status of drowsiness of the driver. Google Cloud Platform is used for training CNN and the trained model is implemented on Raspberry Pi. The trial results demonstrate that the presented technique has reduced training time of CNN network model, high acknowledgment exactness of condition of eyes and can distinguish the drowsiness viably.