Object Detection and Analysis of Human Body Postures Based on TensorFlow
Ling Xie, Xiao Juan Guo · 2019
Human body posture recognition has been an important concern of research in many fields. In this field, a human pose estimation algorithm called OpenPose has been more widely used. But its efficiency is very low. We used deep learning methods based on TensorFlow to recognize human body postures. And we applied it to the judgment of the teacher's teaching states. In order to choose the algorithm that works best for our scenario, we designed eight sets of experimental schemes through combining the classification model and the detection algorithm. Then we used these groups of schemes to detect and classify the teacher's teaching states, including standing, sitting, etc. At last, we analyzed the experimental results in depth and selected the most suitable algorithm for our scenario.