A Fast Eye Tracking Method Based on Convolutional Neural Network

Abstract

The invention discloses a fast human eye tracking method based on a convolutional neural network, comprising the steps of: 1) making a human eye tracking data set and a human eye verification data set; 2) using the human eye tracking data set to train a human eye tracking convolution network , using the human eye verification data set to train the human eye verification convolutional network, the training method is the small batch gradient descent method with momentum, and the loss function is the mean square error; 3) Obtain the initial human eye position in the video; 4) Use the human eye Track the human eye in subsequent video frames by the tracking convolutional network; 5) Use the human eye verification convolutional network to verify whether the human eye tracking result is accurate; 6) If step 5) determines that the human eye tracking result is inaccurate, then re-enter step 3 ) to obtain the initial human eye position; if it is accurate, read the next video frame and return to step 4) to continue tracking the human eye. The present invention breaks through the problems of high error rate and slow speed of the existing tracking method, which cannot deal with complicated real scenes.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a fast human eye tracking method based on a convolutional neural network. Background technique [0002] Eye detection and tracking is a necessary step in face recognition, expression recognition, eye movement analysis, iris recognition, fatigue detection and other technologies, and has extensive application and research significance. [0003] In recent years, the convolutional neural network has made a qualitative improvement in the technical performance of computer vision and image recognition, and the accuracy of computer vision tasks such as target detection, target recognition, and target classification has greatly surpassed previous technologies, even surpassing the human eye. However, there are very few researches and inventions on the use of convolutional neural networks for tracking human eyes. Therefore, this paper invents a convolutional neural network for tra...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art, and propose a fast human eye tracking method based on convolutional neural network, which breaks through the problems of high error rate and slow speed of the existing tracking method, and cannot cope with complex real-life scenes.

Images