Automatic refinement of 3D pose estimation using multiple cameras

The Self-Learning Studio - the project will focus on the automatic refinement of 3D pose estimation using multiple cameras. In contrast to other approaches we will not make use of mocap points or accelerometers, but instead fuse together 3D post estimations generated by single cameras into coherent joint estimates, and use the improved estimations to retain the estimators used by a single camera. It's believed that these bootstrapped estimators will have greater robustness and accuracy than the initial estimators, allowing their use in wider circumstances.