Skip to main content
. 2024 Nov 28;24(23):7618. doi: 10.3390/s24237618

Figure 12.

Figure 12

(a) The variation in training losses seen for three different methods. (b) The variation in training accuracies seen for three different methods. (c) The confusion matrix for CNN using triboelectric voltage signals, where accuracy = 85.4%. (d) The confusion matrix for ResNet18 using triboelectric voltage signals, where accuracy = 96.6%. (e) T-SNE feature visualization for the output features of CNN. (f) T-SNE feature visualization for output features of ResNet18.