Adaptive face identification for small-scale social dynamic environment

Abstract

This article focuses on the problem of modifying the standard face identification approach for use in small-scale social dynamic environments, by focusing on adaptability rather than robustness. A design of adaptive face identification system is presented, along with the employed methods of online learning. The problem of ensuing bias-variance dilemma of an adaptive system is described and solved. The system is shown to be able to aptly adapt to new information and changes the environment, the final classification rate on MUG database was near 99%.