Growing Regression Tree Forests by Classification for Continuous Object Pose Estimation

We propose a novel node splitting method for regression trees and incorporate it into the random regression forest framework. Unlike traditional binary splitting, where the splitting rule is selected from a predefined set of binary splitting rules via trial-and-error, the proposed node splitting method first finds clusters in the training data which at least locally minimize the empirical loss without considering the input space. Then splitting rules which preserve the found clusters as much as possible, are determined by casting the problem as a classification problem. Consequently, our new node splitting method enjoys more freedom in choosing the splitting rules, resulting in more efficient tree structures. In addition to the algorithm for the ordinary Euclidean target space, we present a variant which can naturally deal with a circular target space by the proper use of circular statistics. In order to deal with challenging, ambiguous image-based pose estimation problems, we also present a voting-based ensemble method using the mean shift algorithm. Furthermore, to address data imbalanceness problems present in some of the datasets, we propose a bootstrap sampling method using a sample weighting technique. We apply the proposed random regression forest algorithm to head pose estimation, car direction estimation and pedestrian orientation estimation tasks, and demonstrate its competitive performance.