I am working on an action recognition project and I'm trying to use PoseConv3D as model. I noticed that the model uses the COCO keypoints format(which is the figure below)
while I have the NTU skeletons, which are in this other format.
My question is if there is an easy way to have the approximate location of the eyes and ears keypoints, since there is no information about them in the NTU format. Did anyone encounter this problem and found a solution?
