Compartmentalizing and Ordering Claude Monet's Water Lily Paintings Using a Data-Driven Approach
Open Access
Author:
Hoover, Thomas
Area of Honors:
Data Sciences
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
James Z Wang, Thesis Supervisor John Yen, Thesis Honors Advisor
Keywords:
Data Monet UNET Segmentation Machine Learning convolutional neural network artificial neural network Impressionism
Abstract:
Over the course of his career, the Impressionist painter Claude Monet created over 250 paintings featuring water lilies. However, many of these paintings are difficult to date precisely. The goal of this thesis is twofold. Firstly, this thesis provides a foundation for investigating if it is possible to build an algorithm that generates more precise information about when Monet produced specific works in this series. Secondly, this thesis works to investigate if the use of computer vision can challenge preconceptions about pictoral abstraction within Monet's work. The algorithm created produces a realism score which is computed using UNET segmentation. To create this score, Water Lily paintings are compared to a dataset of over 1900 photographs of water lilies to measure the realism of the paintings. This realism score was computed for 97 Water Lily paintings with dates ranging from 1897-1926. These paintings were selected based on water lily flower profoundness within the individual paintings. Results found were inconclusive for overall abstraction increasing over time, but the sorting of paintings based on their realism score within given date ranges combined with future research in the field may lead to a more narrow timeline for dating these works of art. Secondly, results found may show that Monet's late Water Lily paintings were not as straightforwardly abstract as previously thought.
