Video-Based Human Gesture-Action Mapping

In order to ensure the reliability and timeliness of virtual human posture prediction and skeletal drive, we explore the low-consumption and lightweight virtual human action mapping and posture drive method for real-time video, so as to promote the integration and innovation of virtual human posture prediction and skeletal drive technology. By using depth technology to recognize the transformation relationship of human posture points in space in monocular video. In this paper, we propose a lightweight virtual human posture estimation and action driving method, based on Flask, Unity3D engine and so on, to establish a set of lightweight virtual human limb driving system, which introduces the human body posture action to the client for virtual human limb driving, and reduces to a certain extent the dependence of the virtual human action driving data collection on the equipment and scene.