CNN to predict the angle offset between the eyes focus

Abstract

A convolutional neural network (CNN) is a class of deep, feed-forward artificial neural networks. To construct a CNN model to predict strabismus, eyeball movement data will be collected via recording a video. During the image capturing, interviewees will follow the red dot moving showing on the screen and the camera will record their eyeball movement. The images will be labeled with the degree and the direction. The images with different angle offsets will be used for CNN model as an input dataset. By comparing the direction and position of eyeball, strabismus can be identified and return the angle offset between the eyes focus via the CNN model.