My First Scientific Publication

My paper Demographic Influences on Contemporary Art with Unsupervised Style Embeddings, written with Noa Garcia & Yuta Nakashima, was accepted at the VISART — Vision for Art — workshop at the European Conference on Computer Vision 2020.

The idea was simple but had never been tested at scale: does where an artist comes from, or when they were born, leave a measurable trace in the visual style of their paintings? We scraped tens of thousands of contemporary artworks from online galleries, trained unsupervised style embeddings using a convolutional neural network, and then tested whether demographic variables — nationality, birth decade, gender — could predict clusters in the learned style space. The short answer: yes, especially nationality and birth decade. The effect isn’t huge, but it’s consistent, and it held up across multiple embedding architectures.

This project would not have happened without my time at Osaka University in early 2020. I owe a great deal to Noa Garcia and Yuta Nakashima, who hosted me at the Intelligence and Sensing Lab and the Institute for Datability Science. They gave me the freedom to pursue a strange idea at the intersection of art and machine learning, introduced me to the world of computer vision research, and showed me what it means to work in a truly interdisciplinary lab. Osaka itself — the food, the people, the sheer energy of the city — made those months unforgettable.

More information on this project is available on the project website.