Title:  RAPTOR: Robust Articulated Pose Tracking by Integrating Machine Learning and Model Based Methods

Committee: 

Dr. Vela, Advisor 

Dr. Ames, Chair

Dr. Howard

Abstract: The objective of the proposed research is to develop a method that uses model-based and machine learning methods for pose estimation. Purely machine learning based pose estimation methods are robust to image artifacts but require large annotated datasets and are also unable to account for poses not represented in their training set. On the other hand, model-based methods can emulate an infinite number of poses while respecting the subject's geometry but are susceptible to local minima due to the various artifacts that appear in realistic imaging conditions (e.g. subtle background noise due to shadows or movements). The proposed work outlines how to drive the dataset generation using the same models employed in the model fitting to create a representative training set, and how to include the trained detector's response in the model fitting step to reduce the sensitivity of the method to local minima. When necessary, an intermediate representation is defined so that the two approaches may operate on compatible inputs.  The proposed solution will be applied to articulated pose estimation problems where the accuracy of the estimate is essential to correct interpretation of the motion kinematics.