Research And Application Of Sight Tracking Algorithm Based On Deep Learning

Eye-tracking technology is a hot topic in research in the field of computer vision and human-computer interaction,and has a wide range of application prospects.Therefore,it is of great practical significance to track the eye direction of students in the classroom environment and monitor their listening status in real time.In this paper,on the basis of the original SDM algorithm was improved,improves the classroom environment face positioning accuracy,at the same time combined the technology of deep learning,training field model,and innovative to apply it to the classroom on the video,prints a rise rate curve based on time axis,and according to the rate of rise this metric analysis and guidance is provided for the teaching effect.Firstly,the original SDM algorithm is improved according to the idea of multi-feature selection and feature fusion.SIFT,HOG and LBP features at key points were calculated and pin-two fusion was carried out.The feature combination with minimum estimation error was selected in each iteration.Combined with validation in 300W data set and laboratory environment,the robustness of the improved face alignment algorithm in this paper was proved to ensure the accuracy of face location in classroom environment.Then,the existing public line-of sight estimation data set MPIIGAZE and the residual network ResNet were used to train the line-of sight estimation model.The network is fine tuned on the pre-training model of ImageNet,and the accuracy of the final model set reaches 95% on the test.Finally the application of the algorithm of the early to the classroom listen to on the video through the video frame extraction,face alignment and tailoring,and the student to carry on the division and statistics,and then using eye tracking network is used to estimate the line of sight direction,according to the position information of students and the line of sight direction,according to the statistical strategy for the rise rate,finally according to the output rise rate curve,according to the curve trend of current classroom teaching evaluation and give some guidance advice effect.