Table 4 Summary of K-fold cross-validation over the experimental models (ResNet101, VGG19, and ResNet18) shows accuracy mean and standard deviation recorded among 10-folds, 5-folds, and 3-folds cross-validation for every trained model using SGD and Adam optimizers.
| Â | SA + ResNet101 | SA + VGG19 | SA + ResNet18 | |||
|---|---|---|---|---|---|---|
Mean | STD | Mean | STD | Mean | STD | |
10-Folds | ||||||
 Adam | 19.94% | +/- 7.59% | 52.04% | +/- 26.94% | 27.54% | +/- 9.59% |
 SGD | 34.96% | +/- 24.08% | 22.00% | +/- 7.67% | 88.43% | +/- 6.29% |
5-Folds | ||||||
 Adam | 24.59% | +/- 5.18% | 71.81% | +/- 13.91% | 22.67% | +/- 6.06% |
 SGD | 24.98% | +/- 7.98% | 30.24% | +/- 10.17% | 86.62% | +/- 2.42% |
3-Folds | ||||||
 Adam | 66.02% | +/- 29.23% | 67.18% | +/- 1.97% | 77.22% | +/- 27.65% |
 SGD | 50.32% | +/- 29.67% | 43.50% | +/- 26.16% | 81.85% | +/- 5.50% |
