摘要：The study of mouse social behaviours has been increasingly undertaken in neuroscience research. However, automated quantification of mouse behaviours from the videos of interacting mice is still a challenging problem, where object tracking plays a key role in locating mice in their living spaces. In this talk, we propose a novel method to continuously detect and track several mice andindividual parts without requiring any specific tagging. We evaluate our proposed approach against several baselineson our new datasets, where the results show that ourmethod outperforms the other state-of-the-art approachesin terms of accuracy.
