Table 6 The training and testing accuracy, training time, model size and GFLOPs of 8 models corresponding to citrus images
From: Identification of citrus diseases based on AMSR and MF-RANet
Network | Average training accuracy (%) | Average testing accuracy(%) | Training time | Model size (MB) | GFLOPs |
|---|---|---|---|---|---|
CNN | 65.84 | 68.36 | 10: 14ʹ57ʺ | 0.6 MB | 0.009 |
DenseNet121 | 83.74 | 82.39 | 7: 21ʹ55ʺ | 14.15 MB | 2.8 |
AlexNet | 73.17 | 73.29 | 19: 34ʹ28ʺ | 61 MB | 0.7 |
VGG16 | 81.35 | 83.14 | 8: 00ʹ49ʺ | 138 MB | 15.5 |
NTS-Net | 87.91 | 85.73 | 7: 38ʹ04ʺ | 230 MB | 17.81 |
DFL-Net | 90.56 | 91.75 | 7: 04ʹ24ʺ | 255 MB | 18.6 |
BSNet | 91.44 | 90.07 | 6: 51ʹ32ʺ | 179 MB | 15.03 |
MF-RANet | 97.95 | 96.87 | 3: 47ʹ46ʺ | 24 MB | 6.1 |