Neural nets to detect objects resembling sideshow banners in paintings.

Interdisciplinary collaboration with the Massive Data Institute (MDI).

This project involves the identification of visual evidence within online public art museum collections to further connect circus sideshow banners and religious processional banners beyond their Carnival, Lenten, and Saturnalian historical roots. These red-orange bordered pictorial banners reveal important clues about the social and cultural dynamics of their time, they have often been historically overlooked, they are often only a small detail within a larger artwork (see Panini painting in the Poster Session below), and beyond their connection to specific historical events such as fairs and festivals in the previous centuries, they are often difficult to identify through other search terms without manual inspection. A Machine Learning model can be trained with an advanced filter using neural networks to identify potential banners through a likely combination of Classification + Localization and Object Detection. Thus far, most Object Detection has been applied to photographs and not historical artworks such as paintings and prints. This project offers an exciting and potentially new use of neural networks with the hope of developing an open source tool for other researchers.

MDI graciously funded Chau Nguyen, a second year student in the Data Science for Public Policy Program, for 10 to 15 hours per week during the Spring 2022 semester.