A Generative Study of Sculptural Perception

This study began from a question about how we look at historical sculptures today.

Many public sculptures and religious figures were originally created as representations of important citizens, leaders, or symbolic individuals of their time. However, centuries later, most viewers no longer know who these people were. Their identities gradually become abstract, and the act of viewing becomes inseparable from the imagination of the present.

This project explores how contemporary viewers unconsciously project their own sense of humanity onto historical sculptures.

Using still photographs of sculptures, the work applies AI-based depth estimation to infer three-dimensional structure from a single image. The estimated depth information is then used together with diffusion-based image generation to produce speculative human appearances emerging from stone surfaces and sculptural forms.

The purpose of the work is not to search for a single “correct” reconstruction of the past. Instead, it reflects the idea that every viewer may imagine a different human presence behind the sculpture. In contemporary society, those imagined faces are inevitably shaped by the cultural backgrounds, identities, and visual memories of the people looking at them. A viewer from a different generation, country, or ethnicity may unconsciously project a different person into the same sculpture.

The generated images therefore function less as archaeological restorations and more as mirrors of contemporary perception — revealing how the present imagines the past.

The project asks:

Are we recovering history, or seeing ourselves reflected within it?