Ancestral Apparitions

Ancestral Apparitions uses machine learning to evoke simultaneous feelings of familiarity and distance in a meditation on diasporic identity and more broadly, on the pendulum between collective and individual memory. Thus far, a series of 35 small photographs have been made. For these images, a Generative Adversarial Network is trained on my own family photographs and images from The Studies in Tamil Studio Archives and Society (stars.archive). The ML model is then able to produce its own ghostly "family photographs" which are used to create digital negatives and print images on expired photo paper, returning the images to the original medium of the training data.

I consider this work an earlier iteration for what would become my project I knew that if I walked in your footsteps, it would become a ritual. In Ancestral Apparitions, I was still learning about Generative Adversarial Networks and relying on a process called transfer learning or fine-tuning -- meaning this model is pre-trained on a larger dataset of celebrity faces and then further trained on images from my family and stars.archive. In contrast, the Machine Learning model used in I knew that if I walked in your footsteps, it would become a ritual does not work with a pre-trained model and is only trained on my own family photos and not images from stars.archive. Both works explore notions of collective memory and photography.