Sensory Systems | Theoretical & Computational Neuroscience | Cognitive Neuroscience
The bottlenecks in human letter recognition: a computational model
Avi J. Ziskind, Olivier Hénaff, Yann LeCun, Denis G. Pelli*
*Corresponding author: Denis G. Pelli
Psychology & Neural Science, New York University, New York, New York, USA
F1000Posters 2014, 5: 598 (poster) [English]
Poster [1.30 MB]
Vision Sciences Society Annual Meeting 2014, 16 - 21 May 2014, 56.583
Vision Sciences Society
We have implemented two machine-learning models of object recognition by human observers. Both models capture three hallmarks of human performance that cannot be accounted for by template matching:
<li>spatial frequency channels, </li>
<li>effects of letter complexity. </li>
<li>One model is a Convolutional Neural Network (ConvNet), and the other is a texture statistics model followed by a linear classifier.</li>
With appropriate hyper-parameters and training, both models account for spatial-frequency channels, crowding, and effects of letter complexity.
No relevant competing interests disclosed.
Please note that most posters on this site present work that is preliminary in nature and has not been peer reviewed.
This poster is open access subject to the CC BY Creative Commons 4.0 License