image classification backbone 汇总分析

下面是一个list，可以详细看一下

image_classification = [
['name','top1_acc','top5_acc','size'],
['FixEfficientNet-L2',88.5,98.7,480],
['NoisyStudent/EfficientNet-L2',88.4,98.7,480],
['BiT-L/ResNet',87.54,98.46,-1],
['FixEfficientNet-B7',87.1,98.2,66],
['NoisyStudent/EfficientNet-B7',86.9,98.1,66],
['FixEfficientNet-B6',86.7,98,43],
['FixResNeXt-101 32*48d',86.4,98.0,829],
['NoisyStudent/EfficientNet-B6',86.4,97.9,43],
['FixEfficientNet-B5',86.4,97.9,30],
['NoisyStudent/EfficientNet-B5',86.1,97.8,30],
['FixEfficientNet-B4',85.9,97.7,19],
['KDforAA/EfficientNet-B8',85.8,-1,88],
['FixEfficientNet-B8/MaxUP+CutMix',85.8,-1,87.42],
['FixEfficientNet-B8',85.7,97.6,-1],
['AdvProp/EfficientNet-B8',85.5,97.3,88],
['KDforAA/EfficientNet-B7',85.5,-1,66],
['ResNeXt-101 32*48d',85.4,97.6,829],
['EfficientNet-B8/RandAugment',85.4,-1,-1],
['BiT-M/ResNet',85.39,97.69,928],
['NoisyStudent/EfficientNet-B4',85.3,97.5,19],
['AdvProp/EfficientNet-B7',85.2,97.2,66],
['ResNeXt-101 32*32d',85.1,97.5,466],
['FixEfficientNet-B3',85,97.4,12],
['EfficientNet-B7',85,-1,-1],
['ResNeSt-269',84.5,-1,-1],
['EfficientNet-B7',84.4,97.1,66],
['GPIPE',84.4,97,557],
['ResNeXt-101 32*16d',84.2,97.2,194],
['AssembelResNet152', 84.2,-1,-1],
['ResNeXt-101 32*8d/Semi-weakly sup',84.3,97.2,88],
['TResNet-XL',84.3,-1,77],
['ResNeXt-101 32*16d/semi-weakly sup',84.8,97.4,193],
['NoisyStudent/EfficientNet-B3',84.1,96.9,12],
['EfficientNet-B6',84,96.9,43],
['AmoebaNet-A',83.9,96.9,469],
['ResNeSt-200',83.9,-1,-1],
['FixPNASNet-5',83.7,96.9,86.1],
['FixEfficientNet-B2',83.6,96.9,9.2],
['MultiGrainPNASNet(500px)',83.6,96.7,-1],
['ResNeXt-101 32*4d(semi-weakly sup)',83.4,96.8,42],
['EfficientNet-B5',83.3,96.7,30],
['MultiGrainSENet154(450px)',83.2,-1,-1],
['MultiGrainPNASNet154(450px)',83.1,-1,-1],
['MultiGrainSENet154(400px)',83.0,96.5,-1],
['ResNeSt-101',83,-1,-1],
['Oct-ResNet-152(SE)',82.9,96.3,67],
['PNASNet-5',82.9,96.2,86.1],
['NASNET-A',82.7,96.2,88.9],
['MultiGrainSENet154(500px)',82.7,-1,-1],
['Harm-SE-RNX-101 64*4d',82.66,96.29,88.2],
['FixEfficientNet-B1',82.6,96.5,7.8],
['EfficientNet-B4',82.6,96.3,19],
['MultiGrainPNASNet(400px)',82.6,-1,-1],
['FixResNet-50 Billon',82.5,96.6,-1],
['NoisyStudent(EfficientNet-B2)',82.4,96.3,9.2],
['SCARLET-A4',82.3,96,27.8],
['ResNeXt-101 32*8d',82.2,96.4,88],
['AOGNet-40M-AN',81.87,95.74,-1],
['ResNeXt-101 (cutmix)',81.53,94.97,-1],
['NoisyStudent(EfficientNet-B1)',81.5,95.8,7.8],
['PyConvResNet101',81.49,95.72,42.3],
['DPN-131(320*320)',81.38,95.77,80],
['MultiGrainPNASNet(300px)',81.3,-1,-1],
['DPN-98',81.28,95.6,-1],
['Res2Net-101',81.23,94.43,-1],
['ResNet-50',81.2,-1,-1],
['EfficientNet-B3',81.1,95.5,12],
['DPN-98（320*320）',81.06,95.56,-1],
['DPN-92',80.96,95.47,-1],
['ResNeXt-101 64*4',80.9,95.6,83.6],
['ResNet-200(supervised contrastived)',80.8,95.6,-1],
['DPN-92',80.66,95.34,-1],
['ResNeSt-200(fast AA)', 80.6,95.3,-1],
['NAT-M4',80.5,95.2,9.1],
['FixEfficientNet-B0',80.2,95.4,5.3],
['Inception ResNet V2',80.1,95.1,55.8],
['RandWire-WS',80.1,94.8,-1],
['DPN-131',80.07,94.88,80],
['DPN-98',80.07,94.88,-1],
['ScalaNet-152',79.94,94.82,-1],
['ResNet-200',79.9,95.2,-1],
['NAT-M3',79.9,94.9,9.1],
['RegNetY-8.0GF',79.9,-1,39.2],
['Modified Aligned Xception',79.81,94.83,-1],
['SKNet-101',79.81,-1,48.9],
['CSPResNeXt-50(Mish+Aug)',79.8,95.2,20.5],
['EfficientNet-B2',79.8,94.9,9.2],
['FixResNet-50CutMix',79.8,94.9,-1],
['RegNetY-4.0GF',79.4,-1,20.6],
['LIP-ResNet-101',79.33,94.6,42.9],
['MutiGrain R50-AA-500',79.4,94.8,-1],
['DPN-92',79.27,94.63,-1],
['Xception',79,94.5,22.8],
['MUXNet-xs',66.7,86.8,1.8],
['MobileNetV2',72.56,90.81,3.34],
['MobileNetV3-Large 1.0',75.2,-1,5.4],
['ResNet-50',77.5,-1,29.38],
['ECA/ResNeXt-101',78.6,94.34,42.49],
['Inception V1',69.8,89.9,5],
['MixNet-S',75.8,92.8,4.1],
['MixNet-I',76.6,93.2,4.0],
['MixNet-M',77,93.3,5],
['ResNet-101',78.25,93.95,40],
]

backbone

了解历史Xception， ResNext， 
[VGG,Inception,ResNet, DenseNet, ResNext,Xception,sE,SKnet]
[mobileNetv1,v2,v3,-I-M-S，shufflenet,  ]这些事标杆，模型大小或者精度不比这些更加优秀就不用看了

当前的是efficientnet 还有各种技巧