Pose extraction refers to the process of annotating the body frame of a person in images or videos. While many current technologies can do this, each has their own advantages and disadvantages, for example, some are fast but not accurate, while some are very accurate but also computationally expensive. As we are developing a real-time smart workout assistant that utilizes pose extraction, we require a technology that is balanced in both time and precision. Furthermore, we will compare the chosen one with professional and expensive system of KinaTrax (https://www.kinatrax.com/kinatrax-motion-capture). to This project aims to determine such a technology. For this project, we are:

Installing different existing pose extraction technologies

Test the installed packages on different situations

Provide a report on the packages’ strengths and weaknesses

** *Field experiment on comparing the chosen open-source package with professional KinaTrax

Compare the performance of open source 3D pose extraction packages using different types of exercise videos

*** On site field experiment Canceled due to COVID-19