Accuracy Enhancement in Face-pose Estimation Network Using Incrementally Updated Face-shape Parameters

In this paper, we pursue the refinement of a face-pose estimation method using incrementally updated face-shape parameters. Network-based algorithms generally rely on an on-line training process that uses a large dataset, and a trained network usually works in a one-shot manner, i.e., each input image is processed one by one with a static network. On the other hand, we expect a great advantage from having sequential observations, rather than just single-image observations, in many practical applications. In such cases, the dynamic use of multiple observations can contribute to improving system performance. In our previous study, therefore, we introduced an incremental personalization mechanism using sequential observations to a network-based face-pose estimation method, where the averaged parameters in iterative face-shape estimations are used for face-pose estimation. The head pose estimation accuracy of our method was about 2 degrees. In this paper, we conduct an experiment to examine the error distribution of face-shape estimation and discuss an effective incremental personalization mechanism to update the face-shape parameters based on the error distribution.