Table 3.
The accuracy (%) achieved for the three classifiers trained with fused spatial-spectral features of the multiclass classification category.
Feature set | SVM | DT | NB |
---|---|---|---|
Two spatial-spectral feature sets | |||
DenseNet + Inception | 94.86 | 93.15 | 93.14 |
DenseNet + ResNet-50 | 94.57 | 93.43 | 93.43 |
DenseNet + ResNet-101 | 95.14 | 91.7 | 91.7 |
ResNet-50 + ResNet-101 | 94 | 89.14 | 88.86 |
ResNet-50 + Inception | 96.29 | 92 | 92 |
Inception + ResNet-101 | 94.23 | 89.7 | 89.7 |
Three spatial-spectral feature sets | |||
DenseNet + Inception + ResNet-50 | 96.86 | 94 | 94 |
DenseNet + Inception + ResNet-101 | 96.29 | 93.43 | 93.43 |
DenseNet + ResNet-50 + ResNet-101 | 96.57 | 93.71 | 93.71 |
Inception + ResNet-50 + ResNet-101 | 95.71 | 91.4 | 91.4 |
Four spatial-spectral feature set | |||
DenseNet + Inception + ResNet-50 + ResNet-101 | 96.29 | 94 | 94 |