Ella Lu
North Carolina School of Science and Mathematics
Durham, NC
Compositional Analysis of Visual Art Structure
Ella Lu, 17, of Chapel Hill, built a framework that uses AI to analyze visual art for her Regeneron Science Talent Search computer science project.
View Poster
Humans can intuitively understand art, but computers need a rule-based system to evaluate it. Ella developed Compositional Analysis of Visual Art Structure (CANVAS), a system for analyzing how visual elements are arranged within an artwork. One recognized technique, called steelyard composition, balances a large, eye-catching element on one side of an artwork with smaller elements on the other.
Ella manually checked pieces in a large public dataset of Impressionist landscape paintings for steelyard composition and other techniques, then used this information to train an AI model. She also used Grounding DINO, an existing AI tool that helps computers recognize objects in images. Together, these helped CANVAS identify steelyard composition in paintings. Ella’s work may help machines understand art the way humans do and support large-scale analysis of art.
The child of Jiangang Lu and Shu Lu, Ella attends North Carolina School of Science and Mathematics (Durham).
Beyond the Project
Ella is co-president of her school’s Girls Who Code Club, where she develops and leads weekly lessons on the programming language Python.
FUN FACTS: Ella is editor-in-chief of Blue Mirror, her school’s literary arts magazine. She also draws, paints, crochets, and plays piano and guitar.
