Research on Eye Tracking Technology Combining Deep Learning and Object Recognition Algorithm

In order to make the application of eye tracking technology more convenient and improve its accuracy, this paper proposes an eye tracking method combining deep learning and object recognition algorithm. Firstly, the pupil center coordinates in the complex background environment were obtained by the target recognition algorithm to calculate the direction of the line of sight, then a Temporal Feature Multiplexing Network (TFMN) is designed to correct the target recognition algorithm. In the experiment of the influence of shooting position and continuous eye movement on the performance of the algorithm, the average error rate and variance of the algorithm after correction are improved. The results show that the target recognition correction based on TFMN has high accuracy and stability, expanding the application scenarios of eye tracking technology.