We focus on the challenging problem of efficient mouse 3D pose estimation based on static images, and especially single depth images. We introduce an approach to discriminatively train the split nodes of trees in random forest to improve their performance on estimation of 3D joint positions of mouse. Our algorithm is capable of working with different types of rodents and with different types of depth cameras and imaging setups. In particular, it is demonstrated in this paper that when a top-mounted depth camera is combined with a bottom-mounted color camera, the final system is capable of delivering full-body pose estimation including four limbs and the paws. Empirical examinations on synthesized and real-world depth images confirm the applicability of our approach on mouse pose estimation, as well as the closely related task of part-based labeling of mouse.
View on arXiv