Purchase this article with an account.
Isamu Motoyoshi; Visual aftereffects in natural object categories. Journal of Vision 2013;13(9):62. doi: 10.1167/13.9.62.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We recently showed that adaptation to a realistic 3D object with a particular shape and material alters the appearance of the subsequent object (Motoyoshi, VSS 2012). The aftereffect is robustly induced even by adapting to synthetic stimuli such as band-pass noise, and to stimuli at remote spatial locations (Motoyoshi, ECVP 2012). These findings raise a possibility that the perception of 3D shape and material is based on a neural population of low-level image features represented within a large receptive field at high levels. Using this object aftereffect (OAE), the present study examined whether such image-based coding is also relevant for object categorization, the primary goal of ventral visual processing. Observers were shown with one of 20 morphed photographs between an apple and a pear, and were asked to classify it into 'apple' or 'pear'. The morphing level that gave 50% apple/pear response was defined as her/his categorical boundary between the two. We found that adaptation (4 sec) to one extreme (e.g., apple) caused an apparent shift in the boundary toward the other (e.g., test image at the boundary appeared a pear), even when the test stimulus was presented at non-adapting locations. The aftereffect was also caused, though weakly, by adapting to a texture, synthesized with Portilla-Simoncelli's algorithm, that had similar image statistics with those of apple or pear and was outlined by the either's contour shape. The amount of the aftereffects depended on the combination of texture, contour shape, and category; the adaptor with apple's silhouette and apple-like texture had the largest effect. These results support a notion that a population of low-level features within a large receptive field, which is analogous to 'bag of keypoints' in machine vision, plays a significant role not only in shape and material estimation, but also in object recognition by humans.
Meeting abstract presented at VSS 2013
This PDF is available to Subscribers Only