Motion Capture - Optimization

Abstract

Motion Capture (MC) technology, prevalent in film production and virtual reality gaming, captures and reproduces human skeletal movements. This study seeks to operationalize Motion Capture in practical settings by employing six HiPNUC accelerometer sensors to forecast the motion of 18 SMPL skeleton joints. Our primary task involves harnessing the potential of HiPNUC’s sensors and optimizing the computational engine inspired by Stanford University’s Yifeng Jiang and their work on ”Transformer Inertial Poser.” Our key accomplishment lies in the successful integration of HiPNUC’s accelerometer sensors. We have implemented these sensors, established an Engine ONNX Runtime and streamlined processing times through multiprocessing.