A Deep Learning Approach to the Artwork of Art Museums Utilizing Instagram Hashtags

Abstract

This study is conducted to examine the potential of hashtag (#) images to classify artwork. This study has two aims, Firstly, an attempt to find the reliability of using shareholders’ Instagram images. Secondly, as an approach to predict the style of the art museum with a degree of validity. CNNs (Convolutional Neural Networks) are used for image classification from hashtags that can determine image style from the Instagram hashtag. For data sets, top-rated hashtags from Instagram - #Cute, #Fashion, #Nature, #Food - are firstly processed in the frequency domain to artwork images which can be treated as a hashtag classification. 1,000 painting artwork posts from each art museum – MoMA (Museum of Modern Art) and the Getty (J. Paul Getty Museum) - were collected for test data. As for the artwork image cluster, 8,000 images from hashtags are inputted into CNNs. The output data resulting from the deep learning process is evaluated using a Tensor board. Images of #Cute, #Fashionable, #Food, #Natural trained in a minute. Each label consists of 2,000 Instagram images with over 80% accuracy in the reliability of using hashtag images. This data proves that it works to classify artwork images. #Fashion predicted as MoMA and #Natural predicted as the Getty with a high score of consistency. Finally, compared to MoMA, the Getty has more consistency with the 4 hashtags listed above. Therefore, this study has concluded that hashtags are expected to help shareholders search for artwork and art museums.