From clothes and hairstyles to fashion accessories, humans use a great range of stylistic elements to express themselves, impress others, demonstrate their individualism, or show that they belong to a group. We present new methods to use images as a high-frequency, granular source for the analysis of cultural change. Despite its central importance as a form of social interaction and self-expression, and a rich body of theoretical work, empirical work on style choices is rare. We measure similarity over time and space, tracking the timing and location of influential style innovations. To illustrate our methods, we systematically exploit data from more than 14 million high school yearbook pictures of graduating US seniors to analyze persistence and change in style. We document a striking convergence of male and female style characteristics. This is driven by rising male individualism and declining male persistence across generations from the late 1960s onwards. Also, style polarization increases sharply across commuting zones from the 1970s onwards. In addition, we develop a novel measure of style innovation and show that it predicts patenting by cohorts later in life, suggestive of broader societal trends facilitating innovation across a range of domains. Overall, our results highlight the usefulness of images as a source for cultural economics.