원문정보
초록
영어
deviantART is one of the leading social online network sites with a focus on user-generated artworks. The website has a rich data archive of around 150 million images uploaded by its 15 million members, making it the largest art platform today. This paper describes an open source toolkit that provides a humanities scholar with necessary computational tools to analyse and visualise deviantART and similar art collections. To this end, we combine tools from different research fields such as network analysis, computer vision, machine learning and data visualisation. The toolkit provides the functionality to extract data about members and their artworks directly from the deviantART website, using network analysis to select key members for further investigation. The chosen members’ images are automatically downloaded and annotated with different image features, along with which they can be visualised. The visualisation options offered in the implemented toolkit links images to their originals and can be used to explore and analyse the dataset in an interactive way. The toolkit also features an SVM-based classifier to automatically select features to discriminate artists, artworks and styles, which is hidden from the user behind a simple “suggest features” option.
목차
I. INTRODUCTION
II. METHODOLOGY
III. NETWORK ANALYSIS
IV. IMAGE ANALYSIS
A. Feature extaction
B. Cognitively-inspired features
V. VISUALISATION
A. Related work
B. Visualisation software
C. The deviantART toolkit
VI. EXPERIMENTS
A. The deviantART network
B. Image and classification experiments
VII. CONCLUSIONS
References