Skip to main content
. 2019 Mar;149:188–199. doi: 10.1016/j.isprsjprs.2019.01.015

Fig. 8.

Fig. 8

Example class attention maps of (a) images in UCM multi-label dataset with respect to (b) bare soil, (c) building, (d) car, (e) court, (f) grass, (g) pavement, (h) tree, and (i) water. Red indicates strong activations, while blue represents non-activations. Besides, normalization is performed based on each row for a fair comparison among class attention maps of the same images. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)