Since the coordinates of the annotated joints are labeled in the image space, poses which are very similar to each other might appear in different parts of the image, resulting in input vectors with very different values. We, thus, normalize them using the following procedure: first we find the center of the torso (C T ) by averaging the coordinates of the neck and at least one of the shoulders and hip joints; the pose is then translated to the obtained 2D point and finally scaled by the distance between the neck and C T .
closed as unclear what you're asking by whuber♦ Mar 15 at 14:21
Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.