Isita's Senior Project
Project Goals
My goals are to create photorealistic portraits of Roman women based on sculptures of them and historical information about their physical features. In this project, I aim to learn about utilizing Artificial Intelligence to elevate the realism of my images beyond a mere digital colorization. I also hope to make Roman history, especially the female side, more accessible to anyone on the internet, regardless of classics background.
Photorealistic: Looking like a photo of a real person more so than a sculpture or painting.
Background
Why I Chose This Project
Recently, there has been a push towards using digital art techniques to gain a deeper understanding of history. Some artists and scientists have been utilizing computer science and art techniques to make photo-realistic portraits from sculptures of historical Roman figures, like Augustus. One such project is by Mr. Daniel Voshart (https://voshart.com/roman-emperor-project). Another artist, Haroun Binous from Tunisia, created similar portraits of Roman Emperors inspired by Mr. Voshart’s work (https://www.demilked.com/roman-emperor-recreations-haroun-binous/).
While there is work done on facial reconstruction based on ancient skeletons, such as the portrait of Adelasius Ebalchus, developed by Oscar Nilsson, an archaeological facial reconstructor, there is not as much previous work on portraits based on sculpture (https://www.nationalgeographic.com/culture/article/see-facial-reconstruction-swiss-man-end-roman-empire).
My project seeks to continue this novel work of utilizing sculptures in historical facial reconstruction, but through a different, more diverse lens. The women of history have long been disregarded and forgotten in accounts, and though there is progress, equality and inclusivity are still needed in both the fields of classics and computer science. This project is novel as previous work focuses on royal and prominent Roman men, rather than women, contributing to the field of classics with a new, inclusive analysis of historical figures.
As a woman in STEM and a Latin scholar, I feel that it is my duty to contribute to empowering other women and minorities with my work. As a student and citizen of the world, I want to make sure that our legacy of humanity is as beautifully diverse as we are in reality. This work will contribute to healing the racial and gender-based biases that have been ingrained into historical study and assist in setting a standard of accuracy and inclusivity in history and computational modeling. As aforementioned, facial reconstruction based on sculpture and facial reconstruction utilizing novel machine learning-based softwares are just beginning to be explored, and this project shall ensure that in this pioneering field, diversity is acknowledged and empowered.
My project utilizes Artificial Intelligence and digital art techniques to photo-realize the sculptures of Greco-Roman women. I specifically focus on generating portraits of Livia Drusilla, Agrippina the Younger, and Julia Domna with the tools of Photoshop and Artbreeder, combining standard digital art techniques with novel artificial intelligence technology. Pictures of sculptures and coinage featuring these women, primary source descriptions of the subjects’ facial features, and general information on the physical features of women of the same ethnicity, are used as training data for the software, hypothetically leading to accurate portraits.
Accuracy and Bias: Current
Biases in Current Classics Study and AI
As I was researching, I once came across a Medium article about a chart made on TheApricity listing out Roman Imperial men’s hair and eye colors(https://medium.com/@davieco/were-roman-emperors-blonde-2255ec77d123). Now this chart pictured on the Medium article seems harmless and perhaps even helpful with the cited sources and organized format. However, TheApricity is a known Neo-Nazi forum and much of the information on the charts is actually forcefully mistranslated and stemming from Nazi texts, just to make it look like many of the Roman emperors were fair-haired and fair-eyed. This horrified me deeply. What if I had never known to look at the credibility of the chart and used it to colorize my subjects? It would mean that I too was erasing history and even supporting the white-supremacy of the Nazi movement.
​
Though that was an extreme example, in classics there is a bias towards assuming most notable people of history were white, which is especially incorrect in terms of Rome, which was an empire of great ethnic diversity. In my reconstructions, I try my best to retain the features of my subjects.
​
However, I acknowledge that my work is not perfect and still is influenced by these biases and hope that in future work, more of history will be shown more accurately.
Accuracy and Bias: Historical
Biases in Depiction of My Subjects in Sculpture
Sculptures inherently convey what the artist and patron wish people to see.
​
This means that the depictions of the women I am using are a base may be more idealized than their true features. In the case of men, it was often encouraged to show marks of age in the sculptures to indicate wisdom. On the other hand, emphasizing youthfulness was more important for women.
​
Other parts of the sculpture can indicate what archetype was meant to be represented as well. For example, Livia Drusilla is typically depicted wearing a Palla, which is a Roman veil. It demonstrates her position as a matronly, wise, and virtuous woman.
​
Additionally, despite the fact that the sculptures were once painted, the colors have degraded over time, leaving us with very little true verification of their features. Though there are detailed descriptions for male emperors' features, that does not exist for female empresses.
​
As such, my work can certainly be improved in the accuracy of color, but also inherently represents the desires and rhetoric imbued in these sculptures.
​
​
​