Abstract
Deep neural networks have become a standard framework for image analytics. Besides the traditional applications, such as object classification and detection, the latest studies have started to expand the scope of the applications to include artworks. However, popular art forms, such as comics, have been ignored in this trend. This study investigates visual features for comic classification using deep neural networks. An effective input format for comic classification is first defined, and a convolutional neural network is used to classify comic images into eight different artist categories. Using a publicly available dataset, the trained model obtains a mean F1 score of 84% for the classification. A feature visualization technique is also applied to the trained classifier, to verify the internal visual characteristics that succeed in classification. The experimental result shows that the visualized features are significantly different from those of general object classification. This work represents one of the first attempts to examine the visual characteristics of comics using feature visualization, in terms of comic author classification with deep neural networks.
Highlights
Recent progress in computer vision has facilitated the scientific understanding of artistic visual features in artworks
L and ch Step size step Result: Updated image I∗ L[:, :, :, ch] : the output of channel ch of layer L ; Define the optimization objective: tscore = reduce mean(L[:, :, :, ch]); while not stop condition do Forward: compute activations at ch; Backward: compute gradient w.r.t. image grad ← gradients(tscore, I); Normalize gradient: grad ← grad/std(grad) + 1e−8; Update image: I ← I + grad ∗ step; end Results and discussion This section is dedicated to the experimental results for our two main contributions: comic artist style classification and feature visualization for the classifier
The experimental results are verified in detail, to demonstrate that the classifier could effectively separate the different styles, but made some errors when the styles of different classes were similar
Summary
Recent progress in computer vision has facilitated the scientific understanding of artistic visual features in artworks. Artistic style classification and style transfer are two notable examples of this type of analysis The former aims to classify artworks into one of the predefined classes. The class type can represent the artist, genre, or painting style that effectively represents the aesthetic features of the artwork [1] The latter aims to migrate a style from one image to another [2, 3]. This models a reference image’s statistical features, which are used to transform other images This high-level understanding of visual features enables the effective retrieval, processing, and management of artworks. Both examples have been based on machine learning techniques in recent studies, and deep neural networks in particular. Considering the present influence of popular art forms, investigating the distinguishing aspects of different types of popular artworks would be useful
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call