Stratification of Eye Gaze using the Proposed Convolution Neural Network Model

Humans mostly use their eyes to perceive the outside environment. Human-computer interaction is a typical application for eye gaze monitoring. It is becoming more and more important to classify open and closed eyes for real-time applications such as the detection of driver fatigue or sleepiness to reduce the number of fatal accidents and therefore mortality rates. Additionally, it is possible to forecast some diseases based on how quickly the eyes open and close. The majority of computer vision applications working with pictures face challenges like illumination and resolutions and also must consider accuracy and execution speed in real time. Therefore, the flawless categorization of the condition of the eyes being open or closed forms the objective of the proposed system. In this research, a Convolution Neural Network model is proposed that includes four convolutional blocks consisting of 8 convolution layers, 4 max pool layers,8 Batch Normalization, and 4 dropouts. The model is tuned using 8 batch sizes and Adam optimizer on a total of 20 epochs. The proposed model has obtained the value of accuracy as 0.993 and the loss is 0.0180.