Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 01, 2022 - June 04, 2022
We are studying human and robot ensembles as a highly collaborative work. In this report, we considered the tracking method and the detection method of the performance start point (Einsatz), and obtained the following conclusions. By comparing tracking by cascade, KCF, MIL, and MediaPipe, it was shown that skeleton detection by MediaPipe is desirable comprehensively from the frame rate, tracking stability, and the large number of feature points. In Einsatz detection using MediaPipe and MLP classifier, the optimum parameters were extracted by performing a grid search.
