Thanks @sciencebanshee.bsky.social it was amazing to have you here! Hope to see you soon.

Our work can inform researchers on whether they can substitute traditional infrared-based approaches with video-based ones. It highlights the importance of considering factors such as movement amplitude when evaluating video-based methods, as well as the need for better algorithms and datasets.

OpenPose accuracy varied substantially, ranging from being inaccurate to perfect! Interestingly, this variability reflected the varying amplitude of movements performed - poor estimations for low-amplitude movements while, large-amplitude movements (>10 cm) yielded highly accurate estimates.

Video-based motion-tracking technologies are increasingly being used. But can we rely on such estimations? What factors might determine the accuracy? Using a large dataset of 550+ minutes from 46 subjects, we estimated the accuracy of OpenPose using a traditional infrared system as ground truth.