Representations of Shape in Object Recognition and Long-Term Visual Memory

摘要

Our research has focused on the mechanisms used in visual object recognition.Recent psychophysical results suggest that human perceivers often rely on viewpoint-specific (view-based) representations in conjunction with normalization procedures. Over the past year we have explored the degree to which view-based representations are also appearance based. Specifically we have found that visual recognition across many tasks is sensitive to changes in image properties such as illumination, color, and material (texture). These results indicate that object representations are information rich and that abstract part-based structural-descriptions will not account for much of human recognition performance. Other work has focused on how view-based representations are organized and function across changes in viewpoint. Using 3D stimuli rotated in depth we have investigated the role of task, ranging from basic-level to subordinate-level discriminations, how view-based representations generalize from known members of a class to unfamiliar members of that class, and how perceptual expertise is acquired and influences recognition strategies. We have also been in vestigating the mechanisms used for discriminating between highly similar objects,: e.g., faces.