Fusing Scene Context to Improve Object Recognition