Body Part Localization and Pose Tracking by Using Deepercut Algorithm for King Cobra's BBL (Biting Behavior Learning)

During the recent years, much attention has been given to understand the behavior of king cobra. However, there is no report regarding pose tracking and usage of mathematical model that explain the mechanism of biting dynamics of king cobra. In this study a new method called body part localization technique which is often used in computer vision field is applied for detail behavior learning of king cobra. We explored the method of minimizing the time of direct observation by using deep learning technique. Firstly, we applied the approach of body part localization in order to explore the pose data leading to behavior analysis. Secondly, we formulated a Markov chain model to explain the dynamics of king cobra biting behavior. This model describes the significant potential states of biting behavior by using mathematical model leading to predict the behavioral transition from one state to another. From our investigation we found that king cobra at a normal state will also be irritated if the charmer put external pressure and might have chances of getting injured by 18%. Therefore, even a veteran charmer needs to be very careful while handling king cobra during entertainment show. We also concluded that the body part localization by using body landmarks in freely condition is more effective method to identify the king cobra's body part leading to behavior analysis as the disturbance of its habitation is sharply decreased than that of the repeated direct observation method.