Dataset of images of Roman imperial portraits (2021-2024)

Roman imperial portraits are commonly recognized and identified by relating specific sculptures to commonly-accepted types. The feature that has played a central role in this regard has been hairstyle. Based on the portrayed emperor’s coiffure and (lack of) facial hair, scholars have connected portraits to known classifications and prototypes. However, by focusing mainly on hairstyle, this method runs the risk of excluding portraits that do not adhere to the recognized coiffure repertoire as possible representations of Roman emperors altogether. By using DCNN-based methods, this project investigates the possibilities of digital identification of Roman imperial portraiture at a quantitative level, without using hairstyle. It demonstrates that there are sufficient visual markers in imperial portraits aside from hairstyle, to allow facial recognition software to classify portraits when hairstyle is excluded. The dataset and software codes presented here substantiate our findings, published in the International Journal for Digital Art History. It includes hundreds of photographs of sculptural portraits of Roman emperors from different sources, including personal photography, book scans, and web scrapes.