Text this: A scene invariant convolutional neural network for visual crowd counting using fast-lane and sample selective methods