Python code to detect interaction of medical personnel with the operating table from surveillance videos

Videos of about 40 gynaecological procedures were recorded from a distance in the Leiden University Medical Center.
The purpose was to identify differences in workflow between open, minimally invasive, and robot-assisted surgery.
This code repository was used to analyse human 2D poses extracted from the videos.
It looks at movement speed and position per individual to estimate whether they interact with the operating table or not.
Additionally, it contains code to measure interaction with the patient from Noldus annotation files.