Image Classification on SVHN to MNIST (test)
99.6AccuracyTarget only (baseline)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Target only (baseline)training=Target domain only2019.10 | 99.6 | — | |
| Train on Targetadaptation=fully supervised on target domain2014.09 | 99.51 | — | |
| Target-onlyadaptation=none, trained_on=target domain2016.08 | 99.5 | — | |
| Target-only2016.08 | 99.5 | — | |
| DIRT-T2019.10 | 99.4 | — | |
| Co-DA+DIRT-T2019.10 | 99.4 | — | |
| DTA2019.10 | 99.4 | — | |
| SE*Protocol=MT+CT+TF2019.10 | 98.6 | — | |
| Co-DA2019.10 | 98.3 | — | |
| MCD (Ours)n (number of feature generator updates)=42017.12 | 96.2 | — | |
| MCD (Ours)n (number of feature generator updates)=32017.12 | 95.9 | — | |
| ASSC2017.12 | 95.7 | — | |
| SKDE30Epochs=502022.09 | 95.5 | — | |
| SNeigh30Epochs=502022.09 | 95.5 | — | |
| BFEpochs=502022.09 | 95 | — | |
| SGAN30Epochs=502022.09 | 94.9 | — | |
| VADA2019.10 | 94.5 | — | |
| MCD (Ours)n (number of feature generator updates)=22017.12 | 94.2 | — | |
| BTEpochs=502022.09 | 91.1 | — | |
| SKDE30Epochs=12022.09 | 86.9 | — | |
| CFSMdc=50, beta_M=0.001, beta_C=0.012018.12 | 86.5 | — | |
| ATDAlabeled target samples for validation=true2017.12 | 86.2 | — | |
| Asym. tri-trainingvenue=ICML'172018.12 | 85 | — | |
| Res-paravenue=CVPR'182018.12 | 84.7 | — | |
| BFEpochs=12022.09 | 84.4 | — | |
| DSN w/ DANNadaptation_loss=DANN2016.08 | 82.7 | — | |
| DSNsimilarity loss=DANN2016.08 | 82.7 | — | |
| DSNlabeled target samples for validation=true2017.12 | 82.7 | — | |
| DRCN2017.12 | 82 | — | |
| LETvenue=NIPS'172018.12 | 81 | — | |
| SNeigh30Epochs=12022.09 | 79.2 | — | |
| Joint FTdescription=Model initialized with source pre-train, and fine-tuning on both domains2018.12 | 77.2 | — | |
| Source only (baseline)training=Source domain only2019.10 | 76.5 | — | |
| ADDA2017.02 | 76 | — | |
| ADDA2017.12 | 76 | — | |
| ADDAvenue=CVPR'172018.12 | 76 | — | |
| SGAN30Epochs=12022.09 | 75 | — | |
| Gradient reversal2017.02 | 73.9 | — | |
| Grad. reversalvenue=JMLR'162018.12 | 73.9 | — | |
| DSN w/ MMDadaptation_loss=MMD2016.08 | 72.2 | — | |
| DSNsimilarity loss=MMD2016.08 | 72.2 | — | |
| MMD2016.08 | 71.1 | — | |
| MMD2016.08 | 71.1 | — | |
| MMD2017.12 | 71.1 | — | |
| DANNlabeled target samples for validation=true2017.12 | 71.1 | — | |
| Unsupervised Domain Adaptation by Backpropagationadaptation=unsupervised backpropagation2014.09 | 71.07 | 29.3 | |
| DANNimplementation=replicated2016.08 | 70.7 | — | |
| DANN2016.08 | 70.7 | — | |
| CorReg2016.08 | 69.2 | — | |
| Domain confusion2017.02 | 68.1 | — | |
| Domain confusionvenue=ICCV'152018.12 | 68.1 | — | |
| Source Onlyadaptation=none2017.12 | 67.1 | — | |
| CORAL2016.08 | 63.1 | — | |
| CORAL2016.08 | 63.1 | — | |
| SAmethod_type=Subspace Alignment2014.09 | 61.57 | 5.9 | |
| Source onlydescription=Supervised training on the source only2018.12 | 61 | — | |
| Source only2017.02 | 60.1 | — | |
| Source-onlyadaptation=none, trained_on=source domain2016.08 | 59.2 | — | |
| Source-only2016.08 | 59.2 | — | |
| Source Onlyadaptation=none2014.09 | 59.19 | — | |
| BFEpochs=02022.09 | 33.4 | — | |
| SKDE30Epochs=02022.09 | 31.8 | — | |
| BTEpochs=12022.09 | 20.6 | — | |
| SNeigh30Epochs=02022.09 | 10.7 | — | |
| SGAN30Epochs=02022.09 | 10.4 | — | |
| BTEpochs=02022.09 | 10 | — |