Analysis of Texture Representation in Convolution Neural Network Using Wavelet Based Joint Statistics

We analyze the texture representation ability in a deep convolution neural network called VGG. For analysis, we introduce a kind of wavelet-based joint statistics called minPS that applied to the visual neuron analysis. The minPS consists of 30 dimension features, which come from several types of statistics and correlations. We apply LASSO regression to the VGG representation in order to explain the minPS features. We find that the different scale type cross-correlation does not appear in the VGG representation from the regression weight analysis. Moreover, we synthesize the texture image from the VGG in the context of the style-transfer; we confirm the lack of different scale correlations influences the periodic texture to synthesize.

[1]  Andrew Zisserman,et al.  Very Deep Convolutional Networks for Large-Scale Image Recognition , 2014, ICLR.

[2]  Li Fei-Fei,et al.  Perceptual Losses for Real-Time Style Transfer and Super-Resolution , 2016, ECCV.

[3]  Eero P. Simoncelli,et al.  A functional and perceptual signature of the second visual area in primates , 2013, Nature Neuroscience.

[4]  Michael S. Bernstein,et al.  ImageNet Large Scale Visual Recognition Challenge , 2014, International Journal of Computer Vision.

[5]  Leon A. Gatys,et al.  Image Style Transfer Using Convolutional Neural Networks , 2016, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[6]  Alexei A. Efros,et al.  The Unreasonable Effectiveness of Deep Features as a Perceptual Metric , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[7]  H. Komatsu,et al.  Image statistics underlying natural texture selectivity of neurons in macaque V4 , 2014, Proceedings of the National Academy of Sciences.

[8]  Serge J. Belongie,et al.  Arbitrary Style Transfer in Real-Time with Adaptive Instance Normalization , 2017, 2017 IEEE International Conference on Computer Vision (ICCV).

[9]  Kunihiko Fukushima,et al.  Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position , 1980, Biological Cybernetics.

[10]  D. Hubel,et al.  Receptive fields of single neurones in the cat's striate cortex , 1959, The Journal of physiology.

[11]  Eero P. Simoncelli,et al.  A Parametric Texture Model Based on Joint Statistics of Complex Wavelet Coefficients , 2000, International Journal of Computer Vision.