Basically using Google's Mediapipe blazepose model which is converted by creativIKEP
Thanks to BlazePoseBarracuda we could extract pose features in real time even at smartphones (S20, S21).
Then we needed to use these features to move our avatar asset.
We used method used in ThreeDPoseUnityBarracuda where model updates skeleton directly by calculating difference between previous and current frame's pose landmarks. (Take a close look at the VNectModel)
We modified our code to utillize BlazePoseBarracuda model with VNectModel in ThreeDPoseBarracuda.
