Markerless Motion Capture
Video and deep learning-based alternatives to marker-based motion analysis
Traditional marker-based motion capture is accurate but constrained to the laboratory, requiring specialist equipment, calibration and lengthy setup that limit its use in ecologically valid, field-based, or large-scale settings. This research theme explores markerless approaches - using video and pose-estimation/deep learning methods - to capture human movement without reflective markers or wearable sensors.
The aim is to establish the validity, reliability and practical limitations of markerless methods relative to established marker-based and force-based techniques, and to identify where they can extend biomechanical assessment beyond the lab: pitchside, in clinic, and during real competition.
New projects, collaborations and students in this area are welcome.
