Segmentation, 2025

An installation that deconstructs and recursively reconstructs the human face using machine vision, exploring identity as a shifting, evolving entity.

Segmentation is an embodied immersive installation created for the exhibition Machines as Humans as Machines. Shown through a screen placed in a shop window, the work allows viewers to encounter a segmented, machine‑interpreted version of themselves.

Segmentation is a core method in machine learning: it enables AI systems to break down visual information into structured components. The installation mirrors this process - observing the human face, deconstructing it into discrete fragments, then recursively rebuilding it into a shifting, emergent whole.

This duality reflects segmentation both in technology and biology:

In machine learning: segmentation organizes visual data into meaningful parts, enabling analysis, categorization, and recognition.
In biology: fragmentation is a form of asexual reproduction where a piece of an organism breaks off and grows into a complete, genetically identical individual.

In the installation, identity becomes a fluid, transforming construct rather than a fixed representation. As the system analyzes the viewer, it generates a new self-image that mutates, hides, reveals, camouflages, and redefines itself - questioning traditional boundaries of digital presence, self‑perception, and representation.

The piece is also part of Gáspár Hajdu’s ongoing Digital Therapy series, which explores introspection, self-image, and the psychological dimensions of human–machine interaction.

Technical setup:
4K display • camera • AI/ML segmentation pipeline • real‑time reconstruction

About the exhibition, in Hungarian.

credits
Concept and Creative Coding: Gáspár Hajdu
Special thanks: Melinda Sipos, Sámuel Setényi