Skip to main content
. 2017 Jun 12;17(6):1341. doi: 10.3390/s17061341

Figure 8.

Figure 8

Extracted features from different layers in the proposed “Model 2” network with ordered 8000 sample training images and 7 navigation choices (X-axis: feature dimension; Y-axis: index of samples, best viewed in color). The network nonlinear mappings of fully connected layer 1 (FC1) ((a)→(b)) and FC2 ((b)→(c)) enhance the discriminability one after another.