Towards cognitive and perceptive video systems