Do humans and Convolutional Neural Networks attend to similar areas during scene classification: Effects of task and image type