Implementing Convolutional Neural Networks in Low-Resolution Consumer Imaging Devices for Eye-Gaze Forecasting
Gunjan Sharma, Vatsala Anand, Sheifali Gupta · 2023
For developing technological consumer electronic systems like monitoring systems for drivers and cutting-edge interfaces for users, precise and effective eye gaze estimation is essential. In this process, the eye gaze is located in every frame image or video stream. In this research, a CNN model has been proposed to classify eye-gaze images into four classes. The Eye Gaze dataset has been employed to train the model. The model has been tuned twice by employing Adam and RMSprop optimizer. With both the optimizers the model has outperformed in the classification task. The model has shown a very good accuracy of 96% and a validation accuracy of 90% with Adam Optimizer. The technology in this study can be utilized as an innovative approach in the field of autonomous vehicles and in healthcare by utilizing this approach for medically retarded persons to know their gestures by their eye-gaze.