Developed in collaboration with Ori Blich, RCA (www.oriblich.com), the motive of this project was to help mediate passive interactions in public places. Inspired by the awkward exchange of glances that occurs on the public underground, we decided to approach this idiosyncratic space head-on.

The system was controlled by an arduino which interfaced with a rasberry pi. The computer vision programs ran on the pi owing to its greater processing power.

Face recognition and feature detection were employed to locate the eyes. Having located the eyes, a series of vectors were collected: a vector from a reference point to the mid point between each iris, and the location of the head in space (rotation and translation in 3 dimensions). These eight components were used to train an SVM, classifying the gaze as either 'looking' or 'not looking'.

The final conceptualisation of the physical prototype. The functionality is explained below.