Highly Robust Synthetic Aperture Radar Target Recognition Method Based on Simulation Data Training

Table 9

Recognition results of different CNN methods.

Method	Recognition rate (%)	Params (M)
AlexNet	81.80	57.02
VGG-16 (with BN)	80.08	134.28
VGG-19 (with BN)	80.94	139.59
ResNet-18	78.51	11.18
ResNet-34	84.52	21.29
ResNet-50	80.66	23.51
DenseNet-121	82.81	6.96
DenseNet-161	84.53	26.48
SE-ResNet-18	78.79	11.27
SE-ResNet-34	82.66	21.44
SE-ResNet-50	80.66	26.03
EfficientNet-B0	77.79	4.01
EfficientNet-B1	73.35	6.52
EfficientNet-B3	70.20	10.70
EfficientNetV2-S	82.23	20.18
EfficientNetV2-M	81.52	52.86
RegNetX-800MF	81.80	6.59
RegNetX-3.2GF	84.81	14.29
RegNetY-800MF	81.66	5.65
RegNetY-3.2GF	81.38	17.93