This video was generated by passing individual frames through deep learning network trained to estimate human poses. The inference was run on intel GPU using OpenVino R5 and pre-trained intel models.
This is the original video by Azucena Moreno Conde