List your company
ResearchFormerly WizWand

Image Classification on SVHN 1000 labels (test)

1.89Error RateUDA
-1.151619.379239.9160.4408Jun 20, 2014Sep 8, 2015Nov 27, 2016Feb 16, 2018May 7, 2019Jul 26, 2020Oct 15, 2021
Updated 8mo ago

Evaluation results

MethodLinks
UDA2021.101.89
FixMatch2021.101.96
UDA (RandAugment)Model=WRN-28-2, # Param=1.5M2019.042.23
100% SupervisedBackbone=ConvLarge, labeled_data=100%2019.082.88
MixMatchModel=WRN-28-2, # Param=1.5M2019.042.89
Mean Teacher2021.103.27
DCT with 8 ViewsBackbone=ConvLarge2019.083.29
Triple-GAN-V2Classifier=13-layer CNN [32], Data Augmentation=Standard2019.123.45
ICTModel=WRN-28-2, # Param=1.5M2019.043.53
VAdD (KL) + VATBackbone=ConvLarge2019.083.55
R2-D2Backbone=ConvLarge2019.083.64
MixMatch2021.103.69
VAT+EntMinBackbone=ConvLarge2019.083.86
VAT+EntClassifier=13-layer CNN (similar), Data Augmentation=Standard2019.123.86
SNTGClassifier=13-layer CNN (similar), Data Augmentation=Standard2019.123.86
VAT + EntMinModel=Conv-Large, # Param=3.1M2019.043.86
SNTGModel=Conv-Large, # Param=3.1M2019.043.86
ICTModel=Conv-Large, # Param=3.1M2019.043.89
Mean TeacherBackbone=ConvLarge2019.083.95
MTClassifier=13-layer CNN (similar), Data Augmentation=Standard2019.123.95
Mean TeacherModel=Conv-Large, # Param=3.1M2019.043.95
VAT2021.104.11
VAdD (KL)Backbone=ConvLarge2019.084.16
Temporal EnsemblingBackbone=ConvLarge2019.084.42
MT (our code base)Classifier=13-layer CNN [32], Data Augmentation=Standard2019.124.48
Π-ModelModel=Conv-Large, # Param=3.1M2019.044.82
II modelClassifier=13-layer CNN (similar), Data Augmentation=Standard2019.124.95
ReMixMatch2021.105.16
VATClassifier=13-layer CNN (similar), Data Augmentation=Standard2019.125.42
Shake-Shake (26 2x96d) + AutoAugmentBackbone=Shake-Shake (26 2x96d), Augmentation=AutoAugment2018.055.9
LGA + VATModel=WRN-28-2, # Param=1.5M2019.046.58
FlexMatch2021.106.72
Π-Model2021.107.16
Pseudo-LabelModel=WRN-28-2, # Param=1.5M2019.047.62
Wide-ResNet-28-10 + AutoAugmentBackbone=Wide-ResNet-28-10, Augmentation=AutoAugment2018.058.2
Pseudo-Labeling2021.109.4
Triple-GAN-V2Data augmentation=true, Classifier architecture=13-layer CNN, Number of runs=32019.1210.17
Triple-GAN-V2Data augmentation=false, Classifier architecture=13-layer CNN, Number of runs=32019.1211.08
MTData augmentation=true, Classifier architecture=13-layer CNN, Number of runs=32019.1211.08
Only 1000 labeled imagesBackbone=ConvLarge, labeled_data=1000 samples2019.0811.27
MTData augmentation=false, Classifier architecture=13-layer CNN, Number of runs=32019.1211.68
Shake-Shake (26 2x96d)Backbone=Shake-Shake (26 2x96d), Augmentation=Baseline2018.0512.3
Wide-ResNet-28-10Backbone=Wide-ResNet-28-10, Augmentation=Baseline2018.0513.2
Triple-GAN-V1Data augmentation=false, Classifier architecture=13-layer CNN, Number of runs=32019.1221.44
DCGANclassifier=L2-SVM2015.1122.48
CNN (our code base)Data augmentation=false, Classifier architecture=13-layer CNN, Number of runs=32019.1222.72
SWWAEdropout=true2015.1123.56
SWWAEN=1000, Dropout=true, lambda_L2*=0.42015.0623.56
DADAData augmentation=false, Classifier architecture=13-layer CNN, Number of runs=32019.1224.17
Shake-Shake (26 2x96d)Backbone=Shake-Shake (26 2x96d), Augmentation=Cutout2018.0524.2
Improved-GANData augmentation=false, Classifier architecture=13-layer CNN, Number of runs=32019.1225.47
SWWAEdropout=false2015.1127.83
SWWAEN=1000, Dropout=false, lambda_L2*=0.82015.0627.83
Wide-ResNet-28-10Backbone=Wide-ResNet-28-10, Augmentation=Cutout2018.0532.5
M1+M2Labeled samples=1000, Model type=Generative semi-supervised2014.0636.02
M1+M22015.1136.02
M1+M2N=10002015.0636.02
M1+TSVMFeatures=M1 latent features, Kernel=RBF, Labeled samples=10002014.0654.33
M1+TSVM2015.1154.33
M1+TSVMN=10002015.0654.33
M1+KNNFeatures=M1 latent features, Kernel=RBF, Labeled samples=10002014.0665.63
M1+KNN2015.1165.63
M1+KNNN=10002015.0665.63
TSVMFeatures=Original features, Kernel=RBF, Labeled samples=10002014.0666.55
TSVM2015.1166.55
TSVMN=10002015.0666.55
KNNFeatures=Original features, Kernel=RBF, Labeled samples=10002014.0677.93
KNN2015.1177.93
KNNN=10002015.0677.93