Closed Amirhosein80 closed 3 days ago
Yes, just use forward kinematics. After that, you need to transform the joint positions in the global frame to the root frame by subtracting the root position and left-multiplying it with the inverse root orientation.
Hi, I am currently working on writing a training code for your Pose model. I want to train Pose-S1 and Pose-S2 again, but I don’t know how i can calculate joint positions for training on DIP and Total Capture. Like AMASS can I use forward kinematics or not?