Engilberge, MartinCollins, EdoSüsstrunk, Sabine2017-10-122017-10-122017-10-12201710.1109/ICIP.2017.8296790https://infoscience.epfl.ch/handle/20.500.14299/141361Convolutional 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.text::conference output::conference proceedings::conference paper