| Paper: | PS-2B.41 |
| 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: |
General Shape Features Allow for Categorization of Written Symbols Across Font Variation |
| Manuscript: |
Click here to view manuscript |
| DOI: |
https://doi.org/10.32470/CCN.2018.1215-0 |
| Authors: |
Daniel Janini, Talia Konkle, Harvard University, United States |
| Abstract: |
With extensive experience, humans become experts at recognizing and reading letters and digits. Does the ability to categorize these symbols require a specialized visual feature space, or can this capacity be supported to some extent by a more general feature space also used to represent other visual categories like objects? To examine this question, we tested whether multiple models of general shape features could categorize written symbols across large variations in font. Moderate to robust categorization accuracy was accomplished using deep convolutional neural networks trained to do object categorization, as well as in simpler models like Gist and Normalized Contour Curvature. These models also showed moderate correlations to human classification behavior. Broadly, these results are in line with the possibility that the visual system processes written symbols by leveraging features in place for recognizing real-world objects, rather than primarily relying on symbol-specific feature tuning. |
