Researches Advanced in Human Pose Estimation Based on Deep Learning

Abstract Human pose estimation has always been a popular research topic in the discipline of computer vision, which aims to predict spatial position human body’s key points (parts/joints) from given image or video. Benefiting powerful feature representation ability convolutional neural networks, methods based on deep learning have become mainstream framework this task, obtain results by extracting information at different scales with specific networks and using corresponding process it. This paper, depth study body posture can be divided into single-task multi-person missions. The time method correlation are classified summarized, respectively network architecture, algorithm extraction performance that analyzed. current problems existing field prospect for future prospected.