An Architectural Model for Combining Spatial-Based and Object-Based Information for Attentive Video Analysis.