Glow (manipulating and generating new faces)

Brief description

Users face is detected and captured and the face appears on three screens. Then the user can start manipulating different attributes (mouth, nose, glasses, beard etc.) on their face by using a radio-button interface increasing or decreasing the emphasis on that attribute. The system is using a flow-based generative machine learning model called Glow to generate the images based on a training dataset of a large number of faces.  Adjusting the attributes is similar to what many image filters offers in apps on our phones. The user can create several versions of them selves which are saved on the left hand screen. The system is able to recognize which characteristics could fit with the face of a given person and apparently it tries to identify the gender of the user.

Work that the situation appears in

Title Publication Type Year Creator
Glow Art Diederik P. Kingma, Prafulla Dhariwal, Ars Electronica Futurelab (AT)
Aesthetic characteristics
Machine P.O.V
Not machine P.O.V.

Authored by

UUID
9599bdf9-2609-4ebd-a276-e6c243d0bb12