Adi Anna Telezhynski
Design and Technology track
What does facial recognition technology see when looking at a human face? What elements of that face does it require to identify it as human?
I have created human facial masks at various levels of abstraction. From a certain degree of abstraction, the machines fail to identify faces in these masks. Do we see the subject’s face in all? Can the machine imagine the original face reflected from these masks? I am interested in the gap between the machine’s and our own ability to identify faces. The more the technology develops, the more difficult it will be for us to identify what is uniquely human; we must keep looking for it, to know who we are.
I have studied facial recognition algorithms in depth, and discovered fascinating concepts, such as “ghost face”, blurry and slightly scary faces that represent a group of real faces. This is how the algorithm remembers us: in averages, in groups. Or “face hallucination”, a method for producing high-resolution facial images out of a low-resolution facial image. TO do so, the computer studies all it can about the human face, in order to imagine the original human face out of the blurry photograph.
I use laser scanning to obtain a 3D model of a face, and abstract the scan. I then print the abstracted face on a 3D printer, and apply a facial recognition algorithm to the result. I examine the degree to which the image of the human face can be abstracted, to the point that technology can no longer see a human at all. Then, on this abstracted face, I apply a machine-learning algorithm to reconstruct the original face. I try to look under the hood of the facial recognition algorithms to understand how the machine’s imagination works, whom it imagines it sees. This visual literacy is essential to maintain human subjectivity vis-à-vis the machine.