Grouping Complex Face Parts by Nonlinear Oscillations

摘要

This paper proposes a new neural network model for human face detection. After face parts such as eyes and mouths are extracted from target images in a pre-processing system, they are fed to a recurrent neural network composed of nonlinear oscillations system. The proposed neural network can detect several faces one after another by composing face parts extracted by pre-processing. A whole system works well even when some face parts are lost.