The use of artificial intelligence in GLAM sectors
Artificial intelligence has become a compelling technology applied in GLAM sectors in the recent decade. In cultural heritage institutions, AI could be used for accessing data, analyzing data, and categorizing data to help researchers conduct studies and collaborate with art, history and social science. The computational and quantitative methods are incorporated into the field of digital humanities.
Convolutional Neural Networks is one of the popular technology in visual image studies. Scholars and the GLAM sectors apply this method to digitize paintings and ancient materials; categorize them based on artists/authors, year of origin, styles etc., for preservation and management; establish a database and recommendation system. A convolutional neural network comprises layers of an image, which includes an input layer and hidden layers. The vectors are neurons from each layer, and they are connected with others in the previous layer. The networks can find and analyze relations and correlations between layers and various images, while it is hard to use human eyes to recognize these implicit messages. In this way, the computational method can benefit in investigating, interpreting, and managing ancient works on a larger scale. It will also help find new ways of interaction between humans and computers, humans and artworks, computers and artworks etc.
Educational purpose and benefits
The use of AI in visual image studies empowers learning for non-experts. For students and the general public, it creates greater accessibility with lower cost, high efficiency and digital access. Particularly it is laborious to visit physical museums during the pandemic, not to mention the inequality existing in physical access, such as economic conditions, limitation of cross-boarder travelling or time-wasting. Metadata with interpretation and classification of artworks will support people to look up and acquire resources more effectively. People can find ancient paintings according to various categories, based on keywords, style labels or contents, which is also a chance to study art history and humanities.
This methodology can also help read historical materials written in ancient languages, such as old English and ancient greek, then transcribe physical materials into codes and translate them into modern language. In this case, it undoubtedly breaks language barriers and improves limited access.
Interesting projects such as Google Arts & Culture app relate present photographs and artworks based on likeness. Its principle is also to analyze relations and correlations between vast visual images. From this practical perspective, the technology enhances the connection between the public and the glam sectors while broadening the public’s access to fine arts, historical context, exclusive resources, etc.
Useful tools and related projects
- Artist-Identification-using-CNN
https://github.com/shashankvasisht/Artist-Identification-using-CNN/blob/master/Artist%20ANN.ipynb
It is an artist identification system created by Shashak Vasisht (2018), using Python to classify paintings in vast styles. The names of artists and titles of paintings will be printed out after processing specific features of hundreds of files.
- ArtUK
https://artuk.org
It is a cultural education charity which offers online resources of the public art collection in the UK, including artworks from approximately 3200 institutions. The website provides free digital access for everyone in favour of education and research, and people can search by artists, topics, materials, styles, locations etc.
- Visual Geometry Group
https://www.robots.ox.ac.uk/~vgg/software/
The Department of Engineering Science founds it the University of Oxford provides useful software using Convolutional Neural Networks technology. For example, Image Compare is an offline application processing and comparing the differences between two images; VGG Image Search Engine allows visual searching of a selected area as a query in numerous images.
References
cs231n. (n.d.). CS231N convolutional neural networks for visual recognition. Retrieved May 24, 2022, from https://cs231n.github.io/convolutional-networks/
Messemer, H., Perera, W. L., Heinz, M., Niebling, F., & Maiwald, F. (2020). Supporting Learning in Art History–Artificial Intelligence in Digital Humanities Education. In Workshop Gemeinschaften in Neuen Medien (GeNeMe) 2020. TUDpress.