Shape or Texture: Understanding Discriminative Features in CNNs


TL;DR: We perform an empirical study on the ability of CNNs to encode shape information on a neuron-to-neuron and per-pixel level. To quantify these two aspects, we first approximate the mutual information of latent representations between pairs of semantically related images which allows us to estimate the number of dimensions in the feature space dedicated to encoding shape and texture.



Publication

Md Amirul Islam, Matthew Kowal, Patrick Esser, Sen Jia, Björn Ommer, Konstantinos G. Derpanis, Neil D. B. Bruce
Shape or Texture: Understanding Discriminative Features in CNNs, ICLR, 2021.
[BibTeX] [PDF] 
@InProceedings{islam2021shape,
   title={Shape or texture: Understanding discriminative features in CNNs},
   author={Islam, Md Amirul and Kowal, Matthew and Esser, Patrick and Jia, Sen and Ommer, Bjorn and Derpanis, Konstantinos G and Bruce, Neil},
   booktitle={International Conference on Learning Representations},
   year={2021}
 }

Team

Downloads