The AAGC-LSTM combines spatial and temporal dependency in a single network operation . This is made possible by equipping graph convolutions with adjacency adaptivity, which allows for learning unknown dependencies of the human body joints . Tofurther boost accuracy, we propose longitudinal loss weighting to considernatural movement patterns, as well as body-aware contralateral dataaugmentation . By combining these contributions, we are able to utilize theinherent graph nature of the body, and can thus outperform the state ofthe art for human pose estimation from sparse inertial measurements .

Author(s) : Patrik Puchert, Timo Ropinski

Links : PDF - Abstract

Code :
Coursera

Keywords : graph - body - human - measurements - inertial -

Leave a Reply

Your email address will not be published. Required fields are marked *