Skip to main content
. 2018 Sep 18;18(9):3153. doi: 10.3390/s18093153

Figure 2.

Figure 2

A typical CNN which is composed of an input layer, an output layer, the convolution, normalization, spatial pooling, and fully connected layers. A typical CapsNet consists of an input layer; an output layer; the Conv1d, PrimaryCaps, and DigitCaps layers; a 3-D hyperspectral cube with 200 bands; and a 3-D input patch with size 7 × 7. The 3-D HSI patches are preprocessed for deep learning models. The pixel cubes are fed into the conventional classifiers, i.e., RFs and SVMs.