Image Classification on SYN SIGNS to GTSRB (test)
99.87AccuracyTrain on Target
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Train on Targetadaptation=fully supervised on target domain2014.09 | 99.87 | — | |
| Target-only2016.08 | 99.8 | — | |
| DIRTbackbone=Lipschitz-constraint-based (larger)2019.05 | 99.6 | — | |
| Mean TeacherTraining Protocol=Unsupervised Domain Adaptation, Augmentation Type=TFA, Confidence Thresholding=true2017.06 | 99.37 | — | |
| SupTgtTraining Protocol=Train on Target, Augmentation Type=TFA2017.06 | 99.22 | — | |
| VADAbackbone=Lipschitz-constraint-based (larger)2019.05 | 99.2 | — | |
| Mean TeacherTraining Protocol=Unsupervised Domain Adaptation, Augmentation Type=TF, Confidence Thresholding=true2017.06 | 98.66 | — | |
| SupTgtTraining Protocol=Train on Target, Augmentation Type=TF2017.06 | 98.64 | — | |
| Mean TeacherTraining Protocol=Unsupervised Domain Adaptation, Augmentation Type=TF2017.06 | 98.63 | — | |
| Mean TeacherTraining Protocol=Unsupervised Domain Adaptation, Augmentation Type=minimal, Confidence Thresholding=true2017.06 | 98.53 | — | |
| SupTgtTraining Protocol=Train on Target, Augmentation Type=minimal2017.06 | 98.49 | — | |
| SupSrcTraining Protocol=Train on Source, Augmentation Type=TFA2017.06 | 98.02 | — | |
| ADATraining Protocol=Unsupervised Domain Adaptation2017.06 | 97.66 | — | |
| SupSrcTraining Protocol=Train on Source, Augmentation Type=TF2017.06 | 97.3 | — | |
| SupSrcTraining Protocol=Train on Source, Augmentation Type=minimal2017.06 | 96.95 | — | |
| ATDAlabeled target samples for validation=true2017.12 | 96.2 | — | |
| ATTTraining Protocol=Unsupervised Domain Adaptation2017.06 | 96.2 | — | |
| ATDA2019.05 | 96.2 | — | |
| MCD (Ours)n (number of feature generator updates)=42017.12 | 94.4 | — | |
| MCD2019.05 | 94.4 | — | |
| MCD (Ours)n (number of feature generator updates)=32017.12 | 94 | — | |
| SRDA-Gnoise_method=Gaussian, image_level_optimization=false2019.05 | 93.61 | — | |
| MCD (Ours)n (number of feature generator updates)=22017.12 | 93.5 | — | |
| DSNsimilarity loss=DANN2016.08 | 93.1 | — | |
| DSNlabeled target samples for validation=true2017.12 | 93.1 | — | |
| DSN2019.05 | 93.1 | — | |
| DANN2016.08 | 92.9 | — | |
| DSNsimilarity loss=MMD2016.08 | 92.6 | — | |
| SRDA-Vnoise_method=VAT, image_level_optimization=false2019.05 | 92.44 | — | |
| MMD2016.08 | 91.1 | — | |
| MMD2017.12 | 91.1 | — | |
| DAN2019.05 | 91.1 | — | |
| SRDA-Fnoise_method=FGSM, image_level_optimization=false2019.05 | 90.87 | — | |
| CorReg2016.08 | 90.75 | — | |
| DANNlabeled target samples for validation=true2017.12 | 88.7 | — | |
| DANN2019.05 | 88.7 | — | |
| Unsupervised Domain Adaptation by Backpropagationadaptation=unsupervised backpropagation2014.09 | 88.66 | 56.7 | |
| RevGradTraining Protocol=Unsupervised Domain Adaptation2017.06 | 88.65 | — | |
| Source Only (Lipschitz config)adaptation_protocol=No adaptation, backbone=Lipschitz-constraint-based (larger)2019.05 | 88.6 | — | |
| CORAL2016.08 | 86.9 | — | |
| Source-only2016.08 | 85.1 | — | |
| Source Onlyadaptation=none2017.12 | 85.1 | — | |
| Source Onlyadaptation_protocol=No adaptation2019.05 | 85.1 | — | |
| ASSC2017.12 | 82.8 | — | |
| ASSC2019.05 | 82.8 | — | |
| SAmethod_type=Subspace Alignment2014.09 | 76.35 | 9.1 | |
| Source Onlyadaptation=none2014.09 | 74 | — | |
| SRDA-G*noise_method=Gaussian, image_level_optimization=true2019.05 | 49.86 | — | |
| SRDA-V*noise_method=VAT, image_level_optimization=true2019.05 | 29.04 | — |