Structural Similarity and Spatiotemporal Noise Effects on Learning Dynamic Novel Objects

Abstract

Introduction We live in a dynamic environment. The interplay between our movements relative to other objects and illumination sources produces a continuously changing projection on our retinas. How does our visual system make sense of this visual cacophony to recognize objects? The conventional answer is that the visual system maps dynamic information onto structures that do not vary over time (Marr 1982). For example, popular theories hypothesize that objects are represented as parts and their relations (Biederman 1987; Marr and Nishihara 1978), or as views comprised of visible features (Bu« lthoff and Edelman 1992; Poggio and Edelman 1990; Tarr and Pinker 1989). Recently, however, several studies have underscored the need to understand how the visual system directly uses dynamic information for recognition (

A stereo advantage in generalizing over changes in viewpoint on object recognition tasks

“…An initial set of analyses did not reveal any effects of rotation direction on either RTs or accuracy (but see Stone, 1999;Vuong & Tarr, 2006). We believe that this may be due partially to the lack of self-occlusion, but future studies in this direction are needed.…”

“…As was mentioned above, Experiment 4 consisted of two learning phases followed by a test phase (see Vuong & Tarr, 2006). In the first learning phase, the subjects were shown four target objects.…”

The relative weight of shape and non-rigid motion cues in object perception: A model of the parameters underlying dynamic object discrimination

“…Clearly, there are different types of motion in the environment (see Aggarwal, Ciao, Liao, & Sabata, 1998, for a discussion of different types of motion), but only some types of motion have been systematically investigated with respect to object recognition. Moreover, few studies to date have systematically examined the relative contribution of shape and motion cues to recognition and how these cues interact (e.g., Lander & Bruce, 2000;Newell, Wallraven, & Huber, 2004;Pilz, Thornton, & Bulthoff, 2006;Stone, 1998;Vuong & Tarr, 2006). The evidence from these studies further suggests that observers rely predominantly on shape cues for everyday recognition.…”

Learning to use novel objects: A training study on the acquisition of novel action representations

“…Desmarais, Dixon, & Roy, 2007;Vuong & Tarr, 2006) or by expertise visual training with the objects (Behrmann, Marotta, Gauthier, Tarr, & McKeeff, 2005;Tarr & Cheng, 2003). However, relatively little is known about the effects of direct action experience on object recognition.…”