Evaluation of Interfaces for Describing 3D Faces

M. Phielipp and J.A. Black, Jr. (USA)

Keywords

Face indexing, Face descriptors, 3D faces, human face perception, faces for humancomputer interaction,

Abstract

Face retrieval is a collaborative process between a human user and a retrieval system. To ensure that this process works effectively, it is important that the retrieval system indexes face images using the same criteria as the human user. In other words, it is important that the computer ‘perceive’ human faces in a manner similar to the way humans do. In order to design such a system, it will be necessary to better understand which facial features are salient to humans. To gain a better understanding, by developing some methodology that allows humans to describe faces in an intuitive and efficient manner, while imposing a minimum of bias on the process. This research is aimed at developing and validating such a methodology. We evaluate two alternative computer human interfaces that allow participants to examine and describe 3D representations of faces, as displayed on a 2D LCD screen. The first interface allows the participant to use a mouse to perform translation, rotation, and scaling of the 3D face, while the second plays a video of a face which was captured as the “camera” moved around the 3D face, to capture views from different viewpoints. In general, the participants preferred the mouse-driven interface for studying the faces, and the use of a word list for describing the faces.

Important Links:



Go Back