ABSTRACT ABOUT CALL FOR PARTICIPANTS OUTLINE THE WORKSHOP PARTICIPANTS SCHEDULE SETTING THE SCENE DESIGNS THEMES
Most of the day was spent crafting urban camouflage intended to hide the wearer from the Kinect computer vision system. By the end of the workshop we understood how to dress to avoid detection for the three different Kinect formats.
The Kinect has three different sensing systems - face recognition, skeletal tracking and motion tracking - and participants produced designs that challenged (and defeated) each of these.
Face recognition: The most successful camouflage from face recognition was a projected Rorschach mask where constantly moving patterns of light and shadow prevented the system from identifying facial features. A simpler camouflage for face recognition was to add false features to the face, such as LED eyes, to confuse the system.
Skeletal tracking:In skeletal tracking mode the Kinect seeks to identify the basic human form - head, trunk, two arms and two legs.
To camouflage themselves from this participants tried adding extra body parts. While a second head wasn't successful in fooling the system, an extra leg
did disguise the participant from the skeletal tracking. This approach was developed further into a type of skirt consisting of several extra legs
creating a potentially practical, wearable, camouflage.
Motion tracking: The motion tracking system on the Kinect works by sending out a field of infra-red dots meaning that
disguises that would work for computer vision systems are useless. Instead, participants found they could camouflage themselves from the system
by playing with the depth field of the system using three-dimensional surfaces. The two most effective camouflages for this category were a sheet
of paper cut to create a three-dimensional surface and an umbrella with a foil blanket covering.
