I think moving to `pose_format` would be best. then, it could use similar constants, such as the point names etc. _Originally posted by @AmitMY in https://github.com/sign-language-processing/pose-evaluation/pull/15#discussion_r1955088628_