Convolution Neural Network for Pose Estimation of Noisy Three-Dimensional Face Images

From limited two-dimensional recognition, facial recognition has now been developed to be able to recognize three-dimensional face images, which usually involves process of face pose estimation. As the conventional artificial neural networks has shown low recognition rate to this problem, Convolution Neural Network have been the most potential classifier to determine the pose estimation of a three-dimensional face images. Convolution operation is expected to minimize the effect of distortion and disorientation of the object, and able to efficiently reduce the required parameters. Results show that the CNN system could estimate the pose position of the 3D face images with high recognition rate, however, this recognition rate decline significantly for the noisy buried face images, showing the CNN still need improvement to deal with noisy environments.