Build Results

TOP 1 ACCURACY TOP 5 ACCURACY
MODEL CODE PAPER
ε-REPR
CODE PAPER
ε-REPR
PAPER
Adversarial Inception V3
77.6% -- 93.7% --
DLA-102
78.0% -- 94.0% --
DLA-169
78.7% -- 94.3% --
DLA-34
74.6% -- 92.1% --
DLA-46-C
64.9% -- 86.3% --
DLA-60
77.0% -- 93.3% --
DLA-X-102
78.5% -- 94.2% --
DLA-X-102 64
79.5% -- 94.6% --
DLA-X-46-C
66.0% -- 87.0% --
DLA-X-60
78.2% -- 94.0% --
DLA-X-60-C
67.9% -- 88.4% --
DPN-107
(224x224)
80.2% -- 94.9% --
DPN-107
(320x320, Mean-Max Pooling)
81.8% -- 95.9% --
DPN-131
(224x224)
79.8% 80.1% 94.7% 94.9%
DPN-131
(320x320, Mean-Max Pooling)
81.4% 81.5% 95.8% 95.8%
DPN-68
(224x224)
76.3% 76.4% 93.0% 93.1%
DPN-68
(320x320, Mean-Max Pooling)
78.5% 78.5% 94.4% 94.5%
DPN-68b
(224x224)
77.5% -- 93.8% --
DPN-68b
(320x320, Mean-Max Pooling)
79.4% -- 95.0% --
DPN-92
(224x224)
80.0% 79.3% 94.8% 94.6%
DPN-92
(320x320, Mean-Max Pooling)
81.3% 81.0% 95.7% 95.5%
DPN-98
(224x224)
79.6% 80.0% 94.6% 94.8%
DPN-98
(320x320, Mean-Max Pooling)
81.2% 81.3% 95.7% 95.6%
EfficientNet-B0
76.9% 76.3% 93.2% 93.2%
EfficientNet-B0
(AutoAugment)
76.8% -- 93.2% --
EfficientNet-B1
78.7% 78.8% 94.1% 94.4%
EfficientNet-B1
(AutoAugment)
78.8% -- 94.2% --
EfficientNet-B2
79.8% 79.8% 94.7% 94.9%
EfficientNet-B2
(AutoAugment)
80.1% -- 94.9% --
EfficientNet-B3
(AutoAugment)
81.6% -- 95.7% --
EfficientNet-B4
(AutoAugment)
83.0% -- 96.3% --
EfficientNet-B5
(AutoAugment)
83.7% -- 96.7% --
EfficientNet-B6
(AutoAugment)
84.1% -- 96.9% --
EfficientNet-B7
(AutoAugment)
84.4% -- 96.9% --
EfficientNet-EdgeTPU-L
80.4% -- 95.2% --
EfficientNet-EdgeTPU-M
78.7% -- 94.3% --
EfficientNet-EdgeTPU-S
77.3% -- 93.6% --
Ensemble Adversarial Inception V3
80.0% -- 94.9% --
FBNet-C
75.1% 74.9% 92.4% --
Inception ResNet V2
80.5% 80.1%
95.3% 95.1%
Inception V3
78.8% 78.8% 93.6% 94.4%
Inception V4
80.2% -- 95.0% --
MixNet-L
79.0% 78.9% 94.2% 94.2%
MixNet-M
77.3% 77.0% 93.2% 93.3%
MixNet-S
76.0% 75.8% 92.8% 92.8%
MixNet-XL
80.1% -- 95.0% --
MnasNet-A1
75.5% 75.2% 92.6% 92.5%
MnasNet-B1
74.7% -- 92.1% --
MobileNet V3-Large 1.0
75.6% 75.2% 92.7% --
Modified Aligned Xception
79.6% 79.8% 94.7% 94.8%
NASNet-A Large
82.6% -- 96.0% --
PNASNet-5
82.7% 82.9% 96.0% 96.2%
Res2Net-50 14x8s
78.2% -- 93.8% --
Res2Net-50 26x4s
77.9% -- 93.9% --
Res2Net-50 26x6s
78.6% -- 94.1% --
Res2Net-50 26x8s
79.2% -- 94.4% --
Res2Net-50 48x2s
77.5% -- 93.5% --
Res2Net-DLA-60
78.5% 79.5% 94.2% --
Res2NeXt-101 26x4s
79.2% -- 94.4% --
Res2NeXt-50
78.2% -- 93.9% --
Res2NeXt-DLA-60
78.4% -- 94.1% --
ResNet-101
79.3% -- 94.5% --
ResNet-101-C
79.5% -- 94.6% --
ResNet-101-D
80.4% -- 95.0% --
ResNet-101-S
80.3% -- 95.2% --
ResNet-152
79.7% -- 94.7% --
ResNet-152-C
79.9% -- 94.8% --
ResNet-152-D
80.5% -- 95.2% --
ResNet-152-S
81.0% -- 95.4% --
ResNet-18
70.8% -- 89.8% --
ResNet-26
75.3% -- 92.6% --
ResNet-26-D
76.7% -- 93.2% --
ResNet-34
75.1% -- 92.0% --
ResNet-50
78.5% -- 94.3% --
ResNet-50-C
78.0% -- 94.0% --
ResNet-50-D
79.1% 77.2% 94.5% 93.5%
ResNet-50-S
78.7% -- 94.2% --
ResNeXt-101 32x16d
84.2% -- 97.2% --
ResNeXt-101 32x16d
(288x288 Mean-Max Pooling)
85.0% -- 97.6% --
ResNeXt-101 32x32d
85.1% 85.1% 97.4% 97.5%
ResNeXt-101 32x32d
(288x288 Mean-Max Pooling)
85.9% -- 97.8% --
ResNeXt-101 32x48d
85.4% 85.4% 97.6% 97.6%
ResNeXt-101 32x48d
(288x288 Mean-Max Pooling)
86.1% -- 97.9% --
ResNeXt-101 32x4d
80.3% -- 94.9% --
ResNeXt-101 32x8d
82.7% 82.2%
96.6% 96.4%
ResNeXt-101 32x8d
(288x288 Mean-Max Pooling)
83.5% -- 97.1% --
ResNeXt-101 64x4d
80.6% -- 95.0% --
ResNeXt-50 32x4d
79.4% -- 94.1% --
ResNeXt-50-D 32x4d
79.7% -- 94.9% --
SENet-154
81.2% -- 95.4% --
SENet-154
81.3% 82.7% 95.5% 96.2%
SE-ResNet-101
78.4% -- 94.3% --
SE-ResNet-152
78.7% -- 94.4% --
SE-ResNet-18
71.8% -- 90.3% --
SE-ResNet-34
74.8% -- 92.1% --
SE-ResNet-50
77.6% -- 93.8% --
SE-ResNeXt-101 32x4d
80.2% -- 95.0% --
SE-ResNeXt-101 32x4d
80.9% -- 95.3% --
SE-ResNeXt-101 64x4d
80.9% -- 95.3% --
SE-ResNeXt-26 32x4d
77.1% -- 93.3% --
SE-ResNeXt-50 32x4d
79.1% -- 94.4% --
SE-ResNeXt-50 32x4d
79.9% -- 94.8% --
Single-Path NAS
74.1% 75.0% 91.8% 92.2%
Xception
79.0% 79.0% 94.4% 94.5%