An architecture is described that can utilize photo-sensors embedded in a substantially transparent layer of a UI display for tracking objects that approach or hover over the UI display. The photo-sensors can be configured to detect light of certain wavelengths (e.g., visible light, infrared light) that are propagated toward the UI display, while ignoring light emanating from the UI display when displaying content. Accordingly, by examining various characteristics of the incoming light such as intensity, the architecture can identify a location of a shadow incident upon the display caused by an approaching or hovering selector object blocking portions of incoming light. Additionally or alternatively, the architecture can identify a location of higher intensity for light emanating from the selector object.