Input video: we reproduce the video and we plot a bounding box around the person we want to predict gaze of. Gaze density: we show the gaze density on the selected target view. Gazed area: we show a thresholded version of the gaze density, which allows us to focus on where the person is looking at the moment.
