Table 6.

The classification accuracy (OA (%)) of the linear classification of fused features with PCA transformation, with 60% and 40% of WHU-RS dataset as a training set.

Method	60% of WHU-RS Dataset as a Training Set	40% of WHU-RS Dataset as a Training Set
ResNet50 last conv layer (PCA) + InceptionV3 avg pooling	98.26	95.02
ResNet50 last conv layer (PCA) + Xception avg pooling	97.62	96.52
DenseNet121 conv5_block16_concat (PCA) + Xception avg pooling	97.01	95.69
DenseNet121 conv4_block24_concat (PCA) + Xception avg pooling	97.76	96.68
InceptionV3 mixed_10 (PCA) + ResNet50 avg pooling	96.27	95.85
InceptionV3 mixed_8 (PCA) + ResNet50 avg pooling	98.01	98.67
InceptionV3 mixed_10 (PCA) + Xception avg pooling	96.77	96.02
InceptionV3 mixed_8 (PCA) + Xception avg pooling	98.01	96.35
DenseNet121 conv5_block16_concat (PCA) + ResNet50 avg pooling	98.76	98.34
DenseNet121 conv4_block24_concat (PCA) + ResNet50 avg pooling	96.77	96.52
Xception block14_sepconv2_act (PCA) + DenseNet121 avg pooling	97.51	96.35
Xception block14_sepconv1_act (PCA) + DenseNet121 avg pooling	97.76	96.52
DenseNet121 conv5_block16_concat (PCA) + InceptionV3 avg pooling	96.27	97.51
DenseNet121 conv4_block24_concat (PCA) + InceptionV3 avg pooling	98.01	97.18