Three-dimensional object recognition is viewpoint dependent

Tarr, Michael J.; Williams, Pepper; Hayward, William G.; Gauthier, Isabel

doi:10.1038/1089

Cited by 207 publications

(145 citation statements)

References 12 publications

Supporting

Mentioning

136

Contrasting

Order By: Relevance

“…The results of the second and third experiment are consistent with the notion that haptic pattern recognition may be influenced by viewpoint, and this may be altered by the nature of the stimuli (eg Tarr et al 1998). The visually impaired subjects in experiment 2 all performed better on front and 3-D views than top views.…”

Section: Discussionsupporting

confidence: 79%

“…Some researchers have argued for viewpoint independence (eg Biederman 2001;Biederman and Gerhardstein 1993). Others, notably Tarr and his colleagues (Tarr et al 1995(Tarr et al , 1997(Tarr et al , 1998, have proposed that there may be preferred views, and that this may vary with familiarity and experience, the complexity of forms, and the presence of object details. However, it is not clear which sorts of views might be canonical for touch, since all of us have limited experience with tangible pictures, and many earlyblind persons have none.…”

mentioning

confidence: 99%

See 1 more Smart Citation

The Influence of Viewpoint and Object Detail in Blind People When Matching Pictures to Complex Objects

et al. 2009

View full text Add to dashboard Cite

We examined haptic viewpoint effects in blindfolded-sighted (BS) and visually impaired subjects: early blind (EB), late blind (LB), and very low vision (VLV). Participants felt complex objects and matched tangible pictures to them. In experiment 1, the EB and BS subjects had similar overall performance. Experiment 2 showed that the presence of a detail on the target object lowered performance in the BS subjects, and that matching accuracy was lower overall for top views for the blind subjects. In experiments 3–5, EB, LB, VLV, and BS subjects made judgments about perspective pictures of a model house with more salient object details. In experiment 3, performance was higher for side views than for corner views. Elevated side views were identified more readily than elevated corner views in experiment 4. Performance for top views was higher than for elevated side views in experiment 5, given the relative simplicity of the top-view depictions and salient details. The EB and BS participants had somewhat lower matching accuracy scores than the other groups. We suggest that visual experience is helpful, but not essential for picture perception. Viewpoint effects may vary with experience and object complexity, but the relevant experience need not be specifically visual in nature.

show abstract

Section: Discussionsupporting

confidence: 79%

mentioning

confidence: 99%

The Influence of Viewpoint and Object Detail in Blind People When Matching Pictures to Complex Objects

et al. 2009

View full text Add to dashboard Cite

show abstract

“…First, Experiment 1 showed that with familiar objects depicted as line drawings, RB did not vary across increasing viewpoint disparity between C1 and C2, replicating earlier studies of Dux (2005a, 2005b). Second, Experiments 2-4 showed that RB did vary with viewpoint differences between stimuli under a variety of conditions, but the viewpoint effect was quadratic in shape, in contrast to the linear cost function found in many studies of object recognition (eg., Hayward & Tarr, 1997;Tarr & Pinker, 1989;Tarr, Williams, Hayward, & Gauthier, 1998). The quadratic shape to the RB function suggests a modulation of performance at small and large viewpoint changes, relative to an underlying pattern of viewpoint invariance.…”

Section: Discussion Of Experiments 1-4mentioning

confidence: 99%

Repetition blindness for rotated objects.

Hayward

Zhou²,

Man

et al. 2010

Journal of Experimental Psychology: Human Perception and Perfor

Self Cite

View full text Add to dashboard Cite

Repetition blindness (RB) is the finding that observers often miss the repetition of an item within a rapid stream of words or objects. Recent studies have shown that RB for objects is largely unaffected by variations in viewpoint between the repeated items.In 5 experiments, we tested RB under different axes of rotation, with different types of stimuli (line drawings and shaded images, intact and split), using both novel and familiar objects. Although RB was largely viewpoint invariant, in most experiments, RB was reduced for small (0°) and large (180°) viewpoint differences relative to intermediate rotations. However, these deviations from invariance were eliminated when object images were split, breaking the holistic coherence of the object. These findings suggest that RB is due mainly to the activation of object representations from local diagnostic features, but can be modulated by priming on the basis of view similarity.The human visual system is capable of processing tremendous amounts of information that vary across both spatial and temporal scales. In order to effectively guide the organism around its environment, important information must be detected and then attended to by conscious processes within very short periods of time. A striking phenomenological demonstration of this effect is perception of rapid serial visual presentation (RSVP) streams of objects. For example, participants can successfully detect a target object presented within an RSVP stream in which objects are replaced every 100 ms (Potter & Levy, 1969). This ability requires the fast activation of stored object information, and flexible attentional processes that can update quickly for each presentation.One important constraint on detection of objects within RSVP streams is repetition blindness (RB). If an object is repeated within a single RSVP stream, observers will often report seeing it only once. Thus, while conceptual knowledge about the object is activated, the observer seems unable to keep the two presentations separate in episodic memory. The first demonstration of RB used verbal stimuli, both within seemingly random lists of words and within meaningful sentences (Kanwisher, 1987). In both, repeated words were missed; in the latter context, participants sacrificed sentence grammaticality in failing to report the repeated item. Later experiments used pictorial stimuli instead of (or in conjunction with) verbal materials. For example, Bavelier (1994) investigated RB for both words and pictures. She found significant RB for both types of stimuli, though somewhat reduced in magnitude for pictures. She also found RB across changes in stimulus format, so that if the first critical stimulus (C1) was a picture of an object and the second critical stimulus (C2) was the name of the object, RB was still observed. This result is strong evidence that RB is caused by activation of stimulus information that is not tied to a particular format of presentation, but rather occurs at a deeper conceptual level.Explanations for repetition blindness...

show abstract

“…This property of the sequential-matching task makes it an attractive tool for studying visual representation independently of other knowledge (Biederman & Gerhardstein, 1993;Ellis & Allport, 1986;Hayward & Williams, 2000;Lawson & Humphreys, 1996;Tarr, Williams, Hayward, & Gauthier, 1998) and to distinguish between perceptual and semantic impairments in neuropsychology. For instance, one of the factors used to distinguish patients with apperceptive and associative visual agnosia is their ability to perceptually match or copy objects (Farah, 1990;Humphreys & Riddoch, 1987;Kolb & Whishaw, 1996).…”

Section: Overview Of Experimentsmentioning

confidence: 99%

The Influence of Conceptual Knowledge on Visual Discrimination

Gauthier

James

Curby

et al. 2003

Cognitive Neuropsychology

Self Cite

View full text Add to dashboard Cite

Does conceptual knowledge about objects influence their perceptual processing? There is some evidence for interactions between semantic and visual knowledge in tasks requiring both long-term memory and lexical access. Here we assessed whether similar perceptual/semantic interactions arise during sequential visual matching, a task that does not require access to semantic information. Matching of two-dimensional or three-dimensional novel objects was facilitated when the objects were associated with arbitrarily assigned distinctive artificial semantic concepts as compared to similar semantic concepts. In contrast to prior demonstrations, this effect was obtained in a task that did not require naming objects, and was not affected by participants rehearsing consonant strings, suggesting a direct influence from semantic associations on visual object recognition.

show abstract

Three-dimensional object recognition is viewpoint dependent

Cited by 207 publications

References 12 publications

The Influence of Viewpoint and Object Detail in Blind People When Matching Pictures to Complex Objects

The Influence of Viewpoint and Object Detail in Blind People When Matching Pictures to Complex Objects

Repetition blindness for rotated objects.

The Influence of Conceptual Knowledge on Visual Discrimination

Contact Info

Product

Resources

About