Can we date an artist's work from catalogue photographs?

Abstract

Computer vision has addressed many problems in art, but has not yet looked in detail at the way artistic style can develop and evolve over the course of an artist's career. In this paper we take a computational approach to modelling stylistic change in the body of work amassed by Sir John “Kyffin” Williams, a nationally renowned and prolific Welsh artist. Using images gathered from catalogues and online sources, we use a leave-one-out methodology to classify paintings by year; despite the variation in image source, size, and quality we are able to obtain significant correlations between predicted year and actual year, and we are able to guess the age of the painting within 15 years, for around 70% of our dataset. We also investigate the incorporation of expert knowledge within this framework by consdering a subset of paintings chosen as exemplars by a scholar familiar with Williams' work.