Head Pose Estimation with Improved Random Regression Forests

Figure 2

Illustration of a random regression forest for head poses estimation. For each tree, the binary tests at the nonleaf nodes direct an input sample towards a leaf, where a real-valued, multivariate distribution of the output parameters (i.e., head pose angles) is stored. The forest combines the results of all its trees to produce a probabilistic prediction in the real-valued output space.