Interactive/Generative video installation using face detection.

Today, there are upwards of 270 million official government CCTV cameras across the world, with China being the leading country, comprising over 200 million of the global total (Baltrusaitis, 2020). Hong Kong’s protesters are already dealing with the dystopian order laid down by mass surveillance and facial recognition technologies, hiding their identities in fear of something called “white terror.” When asked, a protester explained - “You will feel it when you post something on Facebook and think you could be accused of being a terrorist” (Hu, 2019).

In “Collective” I explore the aspects of such data collection and guide it towards a creative outcome. Inspired by the first step in the Eigenfaces facial recognition algorithm, I created a script that generates a “mean face” of detections from a hidden camera. I wonder what the face, generated by the collective data of all available CCTV cameras in the world, would look like?

Examples of averaged faces from image search (left) and characters from the Avatar movie (right):

The face detection algorithm sometimes detects non-face objects or forms that resemble faces. This can be fun to play with while trying to generate faces from abstraction.

Averaged faces from false detections in images of: