Skip to main content
. 2018 Apr 7;18(4):1126. doi: 10.3390/s18041126

Figure 5.

Figure 5

Data preprocessing for CNN models. (a) Original hypercube with 1002 channels of various spatial sizes; (b) Subsampled hypercube with 151 channels; (c) Resized to 32 px × 32 px to reduce computation cost; (d) Data augment and normalization.