Hello, can you share the method of producing the saliency maps displayed in Figure 1 of StableNet paper? Look forward to your reply! Thanks!