Technical Program

Paper Detail

Paper: PS-2B.37
Session: Poster Session 2B
Location: Symphony/Overture
Session Time: Friday, September 7, 19:30 - 21:30
Presentation Time:Friday, September 7, 19:30 - 21:30
Presentation: Poster
Publication: 2018 Conference on Cognitive Computational Neuroscience, 5-8 September 2018, Philadelphia, Pennsylvania
Paper Title: Unique contributions of medial axis structure in human object recognition
Manuscript:  Click here to view manuscript
Authors: Vladislav Ayzenberg, Stella Lourenco, Emory University, United States
Abstract: Despite great strides in computer vision, computational models of object recognition remain unable to wholly account for organisms’ ease at identifying objects across changes in view or their ability to categorize novel exemplars. One important feature that is frequently overlooked in these models is the role that an object’s spatial configuration, or ‘structure’, plays in determining object identity and category. Across three experiments, we tested whether a model of structure, based on an object’s medial axis, characterizes human object recognition independently of other models of vision. We found that human perceptual similarity judgments for novel three-dimensional (3D) objects was predicted by the medial axis similarity between objects, and participants preferentially categorized objects by their medial axes across changes to the object’s surface form. Importantly, we found that this pattern of responses could not be accounted for by other visual properties. These results suggest that human object representations incorporate the medial axis and that the medial axis may play a crucial role in object recognition.