Scene Recognition using Visual Attention, Invariant Local Features and Visual Landmarks