A Two-Stream Deep Fusion Framework for High-Resolution Aerial Scene Classification

Table 6

Comparison with the state-of-the-art methods on the AID dataset.

Methods	Training ratios
	20%	50%
BoVW [42]	-	78.66 ± 0.52
MS-CLBP + FV [42]	-	86.48 ± 0.27
GoogLeNet [41]	83.44 ± 0.40	86.39 ± 0.55
CaffeNet [41]	86.86 ± 0.47	89.53 ± 0.31
VGG-VD-16 [41]	86.59 ± 0.29	89.64 ± 0.36
salM^3LBP-CLM [42]	86.92 ± 0.35	89.76 ± 0.45
Fusion by addition [44]	-	91.87 ± 0.36
TEX-Net-LF [43]	90.87 ± 0.11	92.96 ± 0.18
Ours	92.32 ± 0.41	94.58 ± 0.25