Attentive video analysis using spatial-based and object-based cues