Table 1 Properties of our scratch model and eleven pre-trained CNNs
Models | Depth (# layers) | Number of parameters (millions) | Input image size | Complexity | Speed |
|---|---|---|---|---|---|
Scratch-model | 6 | 5.6 | 224 × 224 | Low | High |
AlexNet | 8 | 61 | 227 × 227 | Low | High |
VGG-16 | 16 | 40.1 | 224 × 224 | High | Low |
MobileNet-V2 | 53 | 3.5 | 224 × 224 | Low | High |
ResNet-18 | 18 | 11.4 | 224 × 224 | Low | Low |
ResNet-34 | 34 | 21.5 | |||
ResNet-50 | 50 | 23.5 | |||
ResNet-101 | 101 | 42.5 | |||
ResNeSt-50 | 50 | 22.9 | 224 × 224 | High | Low |
ResNeSt-101 | 101 | 86.74 | |||
Se-ResNeXt | 101 | 28 | 240 × 240 | High | High |
ViT | 24 | 26 | 16 × 16 | High | High |