Table 10.
A comparison of pretrained CNN models.
| Ref. | Model | D | Layer | P (M) | ER | Input size |
|---|---|---|---|---|---|---|
| [89] | AlexNet | 8 | 5 convolution + 3 FC | 60 | 16.4 | 227 × 227 |
| [83] | VGG | 16, 19 | 13–16 convolution + 3 FC | 134 | 7.3 | 224 × 224 |
| [84] | GoogLeNet | 22 | 22 convolution, 9 inception modules | 4 | 6.7 | 224 × 224 |
| [90] | Inception-V3 | 48 | 42 convolution, 10 inception modules | 22 | 3.5 | 229 × 229 |
| [91] | ResNet | 152 | 152 in ResNet-152 | 60.2 | 3.57 | 224 × 224 |
Note: D = depth, ER = error rate, FC = fully connected, M = millions, P = parameter, and all the images used are RGB.