VisGraphNet: A complex network interpretation of convolutional neural features

We propose and investigate the use of visibility graphs to model the feature map of a neural network. Initially devised for studies on complex networks, we employ this type of model for classification of texture images. An alternative viewpoint provided by these graphs over the original data motivates this work. Experiments evaluate the performance of our method using four benchmark databases, namely, KTHTIPS-2b, FMD, UIUC, and UMD and in a practical problem, which is the identification of plant species using scanned images of their leaves. Our method was competitive with other state-of-the-art approaches both in terms of classification accuracy and computational time. Results confirm the potential of techniques used for data analysis in different contexts to give more meaningful interpretation to the use of neural networks in texture classification. (C) 2020 Elsevier Inc. All rights reserved.

Automated summary

Visibility graphs are proposed for modeling the feature map of a neural network, showcasing competitive performance in texture image classification. The experiments confirm the potential of these techniques for providing more meaningful interpretation to the use of neural networks in different contexts.