Kinematic interaction system with improved pose tracking

Techniques are disclosed for providing improved pose tracking for a subject using a 2D camera and generating a 3D image that reproduces the pose of the subject. A 3D skeletal map is estimated from a 2D skeletal map of a subject using, for example, a neural network. A template 3D skeletal map is accessed or generated having bone segments with lengths set based on a given height of the template 3D skeletal map using, for example, anthropometric statistics. An improved 3D skeletal map is then generated by at least relocating one or more of the plurality of bone segments of the estimated 3D skeletal map to more closely match corresponding template bone segments of the template 3D skeletal map. The improved 3D skeletal diagram may then be animated in various ways (e.g., using various skins or graphics) to track the respective motion of the subject. 公开了技术，用于对使用2D相机的主体提供改进的姿态跟踪，并且生成再现主体的姿态的3D图像。使用例如神经网络从主体的2D骨架图估计3D骨架图。访问或生成模板3D骨架图，其具有骨段，这些骨段具有使用例如人体测量统计基于模板3D骨架图的给定高度来设置的长度。然后通过至少重定位估计的3D骨架图的多个骨段中的一个或多个骨段，以更紧密地匹配模板3