Table 2.
OA (%) and SD of different methods under 5:5 and 2:8 training ratios on the AID dataset.
| Method | OA (TP5:5) | OA (TP2:8) |
|---|---|---|
| GoogLeNet [36] | 86.39 ± 0.55 | 83.44 ± 0.40 |
| VGG-16 [36] | 89.64 ± 0.36 | 86.59 ± 0.29 |
| salM3 LBP-CLM [37] | 89.76 ± 0.45 | 86.92 ± 0.35 |
| Fusion by addition [38] | 91.87 ± 0.36 | / |
| Fusion by concatenation [38] | 91.86 ± 0.28 | / |
| TEX-Net-LF [39] | 92.96 ± 0.18 | 90.87 ± 0.11 |
| Multilever Fusion [40] | 94.17 ± 0.32 | / |
| MF2Net [41] | 94.84 ± 0.27 | 91.34 ± 0.35 |
| S-CNN [42] | 95.24 ± 0.18 | 92.38 ± 0.13 |
| RAN [43] | 93.66 ± 0.28 | 92.18 ± 0.42 |
| VGG-16-CapsNet [44] | 94.74 ± 0.17 | 91.63 ± 0.19 |
| DS-CapsNet | 95.58 ± 0.15 | 93.01 ± 0.16 |