Human Pose Detection Based on Multi-scale and Multi-stage Structure Network

Human pose detection is essential for human behaviour recognition. In this paper, an end-to-end network is proposed to improve the precision of pose detection by fusing multi-scale features, detected parts of the human body at coarse stages and the detected human positions. The network parameters are trained uniformly by Adam optimization algorithm. A large quantity of experiments is carried out on MPII dataset and the images captured by ourselves to explore the relationship between the detection rate and the network structure. The experimental results show that the network can achieve a high detection rate.