| This project incorporates soft geometric constraints into tracking algorithms using tools from statistical machine learning. Soft geometric constraints are more robust to the radical appearance changes that occur during occlusions. Tracking performance is measured against ground truth containing rigid pose information (position and/or orientation). |