Skip to content

Performance on ImageNet validation set

Luigi edited this page Oct 4, 2018 · 16 revisions

Accuracy on validation set (single model)

Results were obtained using (center cropped) images of the same size.

Model Version Acc@1 Acc@5
NASNet-A-Large Tensorflow 82.69 96.16
NASNet-A-Large Our porting 82.50 95.45
InceptionResNet-v2 Tensorflow 80.40 95.30
SENet154 Caffe 81.32 95.53
SENet154 Our porting 81.32 95.45
InceptionResNet-v2 Our porting 80.28 95.14
SE-ResNeXt101_32x4d Our porting 80.28 95.02
Inception-v4 Tensorflow 80.20 95.30
SE-ResNeXt101_32x4d Caffe 80.19 95.04
Inception-v4 Our porting 80.10 94.89
ResNeXt101_64x4d Torch7 79.60 94.70
DualPathNet131 Our porting 79.44 94.60
DualPathNet98 Our porting 79.23 94.49
SE-ResNeXt50_32x4d Our porting 79.11 94.48
SE-ResNeXt50_32x4d Caffe 79.03 94.46
Xception Keras 79.00 94.50
ResNeXt101_64x4d Our porting 78.98 94.26
ResNeXt101_32x4d Torch7 78.80 94.40
Xception Our porting 78.79 94.26
SE-ResNet152 Caffe 78.66 94.46
SE-ResNet152 Our porting 78.64 94.39
SE-ResNet101 Our porting 78.42 94.17
SE-ResNet101 Caffe 78.25 94.28
ResNet152 Pytorch 78.25 93.98
ResNeXt101_32x4d Our porting 78.22 93.94
FBResNet152 Torch7 77.84 93.84
SE-ResNet50 Caffe 77.63 93.64
SE-ResNet50 Our porting 77.61 93.80
Inception-v3 Pytorch 77.50 93.59
FBResNet152 Our porting 77.44 93.54
ResNet101 Pytorch 77.31 93.56
DenseNet161 Pytorch 77.15 93.60
DenseNet201 Pytorch 76.93 93.39
CaffeResnet101 Caffe 76.40 92.90
CaffeResnet101 Our porting 76.11 92.70
ResNet50 Pytorch 76.01 92.93
DualPathNet68 Our porting 75.95 92.78
DenseNet169 Pytorch 75.63 92.81
DenseNet121 Pytorch 74.47 91.97
VGG19_BN Pytorch 74.22 91.85
NASNet-A-Mobile Our porting 74.10 91.78
NASNet-A-Mobile Tensorflow 74.00 91.60
BNInception Our porting 73.48 91.55
VGG16_BN Pytorch 73.48 91.54
ResNet34 Pytorch 73.27 91.43
VGG19 Pytorch 72.36 90.85
MobileNet-v2 Pytorch 71.81 90.41
VGG16 Pytorch 71.63 90.37
VGG13_BN Pytorch 71.62 90.36
VGG11_BN Pytorch 70.41 89.72
VGG13 Pytorch 69.98 89.31
ResNet18 Pytorch 69.64 88.98
MobileNet-v1 Pytorch 69.52 88.98
VGG11 Pytorch 68.87 88.66
ShuffleNet Pytorch 67.41 87.26
GoogLeNet Our porting 66.45 87.52
SqueezeNet-v1.1 Pytorch 58.18 80.51
SqueezeNet-v1.0 Pytorch 58.00 80.49
Alexnet Pytorch 56.62 79.06