Color representation in deep neural networks

Martin Engilberge, Edo Collins, Sabine SüsstrunkICIP 2017 · OpenReview

Abstract

Convolutional neural networks are top-performers on image classification tasks. Understanding how they make use of color information in images may be useful for various tasks. In this paper we analyze the representation learned by a popular CNN to detect and characterize color-related features. We confirm the existence of some object- and color-specific units, as well as the effect of layer-depth on color-sensitivity and class-invariance.
BibTeX
@inproceedings{engilberge2017color,
  author = {Martin Engilberge and Edo Collins and Sabine Süsstrunk},
  title = {{Color representation in deep neural networks}},
  booktitle = {ICIP},
  year = {2017},
  url = {https://openreview.net/forum?id=aTXaZLO4E833},
}

Loading related content...

Figure