Table 7 FLOPs and parameter counts in each method.
Detector | Architecture | FLOPs (G) | Parameters (M) |
|---|---|---|---|
Faster R-CNN | ResNet-101 | 253.7 | 61.2 |
ResNet-152 | 302.5 | 76.8 | |
MobileNetv3 | 72.4 | 14.9 | |
SSD500 | ResNet-101 | 149.8 | 34.7 |
ResNet-152 | 171.3 | 48.5 | |
MobileNetv3 | 49.6 | 10.1 | |
R-FCN | ResNet-101 | 188.7 | 42.2 |
ResNet-152 | 222.6 | 56.9 | |
MobileNetv3 | 59.8 | 11.3 | |
YOLOv5 | CSPDarknet53s | 35.8 | 7.3 |
CSPDarknet53x | 54.2 | 18.4 | |
YOLOv8 | Swin Transformer | 64.7 | 23.6 |
PVT | 60.2 | 20.4 | |
Axial Transformer | 57.9 | 19.7 | |
YOLOv10 | CSPNet | 61.7 | 21.8 |
ConvNeXt | 59.4 | 23.5 | |
EfficientNet | 45.3 | 16.2 | |
ViT | 84.6 | 28.9 | |
Swin Transformer | 67.8 | 24.1 | |
PVT | 65.9 | 23.2 |
