What You See Is What You Get: Drawing Inferences From Campaign Imagery