Openpose: Output Normalizing Explanation

[alrzmshy]

In the ouput.md it is mentioned that:

1. An array pose_keypoints_2d containing the body part locations and detection confidence formatted as x1,y1,c1,x2,y2,c2,.... The coordinates x and y can be normalized to the range [0,1], [-1,1], [0, source size], [0, output size], etc., depending on the flag keypoint_scale, while c is the confidence score in the range [0,1].

Could you please explain how the normalization for [0,1] and [-1,1] is done? In the case of [0,1] is it simply division by the height/width of the frame?

Thanks in advance.

[gineshidalgo99]

Yes, with respect to the image size, (0,0) would be top-left, (1,1) bottom-right of the image. And yes to your last question. And having [0,1], [-1,1] = [0,1] * 2 - 1.