Table 5 The robustness of evaluated DCNN architectures against images transformations
From: Plant diseases detection with low resolution data using nested skip connections
Dataset | Architecture | Conditions | |||||
|---|---|---|---|---|---|---|---|
GauBlr | MedBlr | Rot90 | Rot180 | Sc32 | Sc48 | ||
Plantvillage | AlexNet | 62.83 | 65.76 | 79.73 | 81.06 | 86.06 | 93.49 |
VGG | 62.46 | 66.63 | 77.37 | 78.83 | 81.54 | 91.98 | |
MobileNet | 46.66 | 48.12 | 51.98 | 55.29 | 57.37 | 61.18 | |
ResNet | 65.44 | 66.52 | 75.72 | 64.51 | 89.05 | 91.12 | |
Xception | 63.16 | 65.49 | 75.78 | 64.77 | 87.65 | 92.04 | |
ComNet | 72.49 | 74.00 | 84.20 | 85.71 | 93.45 | 95.90 | |
Tea | AlexNet | 57.54 | 63.83 | 58.38 | 53.34 | 74.31 | 78.22 |
VGG | 54.68 | 60.54 | 62.08 | 60.57 | 71.75 | 78.57 | |
MobileNet | 34.77 | 36.45 | 30.50 | 32.13 | 38.00 | 37.84 | |
ResNet | 57.23 | 62.84 | 46.33 | 42.54 | 67.67 | 69.88 | |
Xception | 56.72 | 64.83 | 48.66 | 46.79 | 69.44 | 71.46 | |
ComNet | 66.56 | 75.08 | 62.21 | 60.09 | 83.20 | 85.40 | |