de.mpg.escidoc.pubman.appbase.FacesBean
English
 
Help Guide Disclaimer Contact us Login
  Advanced SearchBrowse

Item

ITEM ACTIONSEXPORT

Released

Poster

Analog topological transformations in basic level object recognition

MPS-Authors
http://pubman.mpdl.mpg.de/cone/persons/resource/persons83944

Graf,  M
Department Human Perception, Cognition and Action, Max Planck Institute for Biological Cybernetics, Max Planck Society;

Locator
There are no locators available
Fulltext (public)
There are no public fulltexts available
Supplementary Material (public)
There is no public supplementary material available
Citation

Graf, M. (2001). Analog topological transformations in basic level object recognition. Poster presented at First Annual Meeting of the Vision Sciences Society (VSS 2001), Sarasota, FL, USA.


Cite as: http://hdl.handle.net/11858/00-001M-0000-0013-E175-8
Abstract
Purpose: Several studies suggest that the image-based approach to object recognition can be extended to basic level recognition, but it is not clear yet how basic level recognition is achieved. It seems that the shapes of category members on the basic level can be aligned by rather simple topological (warping) transformations — for most biological and many artifact categories. This study was motivated by the hypothesis that basic level recognition involves analog topological transformation processes — in order to achieve an alignment of stimulus representation and object representation. Two research questions were investigated: (1) Does categorization performance depend on the amount of topological transformation? (2) If systematic effects can be detected, are these effects caused by analog transformation processes? Methods: Images from 16 biological and 14 artifact categories were scanned. On the basis of two members of a category, with morphing software new exemplars were produced as intermediate morphs at specific transformational distances. In every trial two backward masked images were presented sequentially (sequential matching task). The transformational distance between the two images was varied. Subjects had to decide by key press whether both images belonged to the same basic level category or not. Results: The predictions were confirmed by the data: Both RTs and error rates increased with increasing amount of topological transformation between two objects of the same category. Furthermore, RTs are sequentially additive, which is evidence that the topological transformation traverses intermediate points in the transformational path. Conclusions: Taken together, the data indicate that basic level recognition relies on time-consuming and error-prone topological transformation processes, which have an analog nature. The results support a model of basic level recognition that is based on analog (image-like) representations and analog transformation processes.