Making measurements of how a person moves (for instance, when walking naturally, when practicing a sport, when recovering from injury) is a valuable tool for researchers, medial practitioners and animators. The measurements can help identify injury risks, optimise athlete performance, diagnose the source of aches and pains, monitor the progress of a patient's recovery from injury, or help an animator turn an actor into a superhero.
The way that motion is measured often involves dressing a person in special clothing, fixing special reflective markers to their bodies, and watching them through specialised camera systems. This can be invasive, awkward, limiting and time consuming to set up, but the marker tracks can be highly accurate and very quick to capture. To make getting the measurements easier, researchers have been trying to remove the need for markers, training computers to better detect and identify parts and points on the human body, to create "markerless" motion capture systems.
The BioCV dataset provides both video imagery and traditional motion capture measurements so that the performance of these "markerless" systems can be compared to the more traditional approach, and identify what level of accuracy is possible, and thus what applications they can be applied to.