Multi-class classification on PACS (test)
82.73Accuracy (Art Painting)Baseline
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| BaselineBackbone=ResNet18, Data Variant=Stylized, alpha=1, p=0.92021.01 | 82.73 | 77.97 | 94.95 | 81.61 | 84.32 | |
| Epi-FCRBackbone=ResNet18, Data Variant=Original2021.01 | 82.1 | 77 | 93.9 | 73 | 81.5 | |
| OursBackbone=ResNet-18, K=22019.11 | 81.28 | 77.16 | 96.09 | 72.29 | 81.83 | |
| DG-MMLDBackbone=ResNet18, Data Variant=Original2021.01 | 81.28 | 77.16 | 96.06 | 72.29 | 81.83 | |
| Mixup (feature-level)Backbone=ResNet18, Data Variant=Mixup2021.01 | 81.2 | 76.41 | 96.31 | 69.67 | 80.9 | |
| DDAIGBackbone=ResNet18, Data Variant=Stylized, alpha=1, p=0.92021.01 | 81.02 | 78.75 | 95.07 | 79.67 | 83.63 | |
| DG-MMLDBackbone=ResNet18, Data Variant=Stylized, alpha=1, p=0.92021.01 | 80.85 | 77.1 | 95.11 | 77.69 | 82.69 | |
| Epi-FCRBackbone=ResNet18, Data Variant=Stylized, alpha=1, p=0.92021.01 | 80.68 | 78.87 | 92.5 | 76.57 | 82.15 | |
| OursBackbone=ResNet-18, K=42019.11 | 80.07 | 75.06 | 95.73 | 74.21 | 81.26 | |
| Pre-trained + EdgeMACTraining source=All 3 other domains, Fusion=Concatenation2017.09 | 80 | 68.7 | 93.7 | 62.7 | 76.2 | |
| OursBackbone=ResNet-18, K=32019.11 | 79.64 | 76.75 | 95.86 | 71.22 | 80.87 | |
| RotationBackbone=ResNet18, Data Variant=Stylized, alpha=1, p=0.92021.01 | 79.51 | 79.93 | 93.55 | 82.01 | 83.75 | |
| JiGenBackbone=ResNet-182019.11 | 79.42 | 75.25 | 96.03 | 71.35 | 80.51 | |
| DDAIGBackbone=ResNet18, Data Variant=Original2021.01 | 79.41 | 74.81 | 95.22 | 69.29 | 79.68 | |
| Deep AllBackbone=ResNet-18, Reference=Current Paper Baseline2019.11 | 78.34 | 75.02 | 96.21 | 65.24 | 78.7 | |
| RotationBackbone=ResNet18, Data Variant=Original2021.01 | 78.16 | 76.64 | 95.57 | 72.2 | 80.64 | |
| Mixup (pixel-level)Backbone=ResNet18, Data Variant=Mixup2021.01 | 78.09 | 71.08 | 93.85 | 66.58 | 77.4 | |
| Pre-trained (RGB)Training source=All 3 other domains2017.09 | 78 | 68 | 94.4 | 47.1 | 71.9 | |
| Deep AllBackbone=ResNet-18, Reference=from D-SAM paper2019.11 | 77.87 | 75.89 | 95.19 | 69.27 | 79.55 | |
| Deep AllBackbone=ResNet-18, Reference=from JiGen paper2019.11 | 77.85 | 74.86 | 95.73 | 67.74 | 79.05 | |
| D-SAM*Backbone=ResNet-18, Uses domain labels=true2019.11 | 77.33 | 72.43 | 95.3 | 77.83 | 80.72 | |
| BaselineBackbone=ResNet18, Data Variant=Original2021.01 | 77.28 | 73.89 | 95.83 | 67.01 | 78.5 | |
| Pre-trained + EdgeMACTraining source=Photo (P), Fusion=Concatenation2017.09 | 73.3 | 34 | — | 27.6 | — | |
| Pre-trained (RGB)Training source=Photo (P)2017.09 | 72.5 | 33.3 | — | 24.8 | — | |
| BaselineBackbone=AlexNet, Data Variant=Stylized, alpha=1, p=0.752021.01 | 71.96 | 72.47 | 88.34 | 76.47 | 77.31 | |
| RotationBackbone=AlexNet, Data Variant=Stylized, alpha=1, p=0.752021.01 | 71.74 | 73.39 | 89.22 | 75.98 | 77.59 | |
| Pre-trained (RGB)Training source=Cartoon (C)2017.09 | 71.7 | — | 86.8 | 37 | — | |
| Siamese (RGB)Training source=All 3 other domains2017.09 | 71.5 | 64.3 | 85.1 | 56 | 69.2 | |
| DG-MMLDBackbone=AlexNet, Data Variant=Stylized, alpha=1, p=0.752021.01 | 70.5 | 70.84 | 88.43 | 75.39 | 76.29 | |
| DDAIGBackbone=AlexNet, Data Variant=Stylized, alpha=1, p=0.752021.01 | 69.35 | 71.1 | 87.7 | 70.99 | 74.79 | |
| Pre-trained + EdgeMACTraining source=Cartoon (C), Fusion=Concatenation2017.09 | 69.3 | — | 85 | 55.3 | — | |
| OursBackbone=AlexNet, K=32019.11 | 69.27 | 72.83 | 88.98 | 66.44 | 74.38 | |
| DG-MMLDBackbone=AlexNet, Data Variant=Original2021.01 | 69.27 | 72.83 | 88.98 | 66.44 | 74.38 | |
| OursBackbone=AlexNet, K=42019.11 | 68.84 | 72.53 | 88.75 | 65.9 | 74.01 | |
| Deep AllBackbone=AlexNet, Reference=Current Paper Baseline2019.11 | 68.09 | 70.23 | 88.86 | 61.8 | 72.25 | |
| JiGenBackbone=AlexNet2019.11 | 67.63 | 71.71 | 89 | 65.18 | 73.38 | |
| Mixup (feature-level)Backbone=AlexNet, Data Variant=Mixup2021.01 | 67.04 | 69.1 | 88.88 | 55.4 | 70.11 | |
| OursBackbone=AlexNet, K=22019.11 | 66.99 | 70.64 | 89.35 | 67.78 | 73.69 | |
| BaselineBackbone=AlexNet, Data Variant=Original2021.01 | 66.83 | 70.85 | 89.78 | 59.75 | 71.8 | |
| Deep AllBackbone=AlexNet, Reference=from JiGen paper2019.11 | 66.68 | 69.41 | 89.98 | 60.02 | 71.52 | |
| MLDGBackbone=CNN2017.10 | 66.23 | 66.88 | 88 | 58.96 | 70.01 | |
| MLDG*Backbone=AlexNet, Uses domain labels=true2019.11 | 66.23 | 66.88 | 88 | 58.96 | 70.01 | |
| Mixup (pixel-level)Backbone=AlexNet, Data Variant=Mixup2021.01 | 66.03 | 68 | 88.9 | 51.18 | 68.53 | |
| Siamese (RGB)Training source=Photo (P)2017.09 | 66 | 38 | — | 31.9 | — | |
| RotationBackbone=AlexNet, Data Variant=Original2021.01 | 65.66 | 71.89 | 89.88 | 62.15 | 72.39 | |
| Epi-FCRBackbone=AlexNet, Data Variant=Stylized, alpha=1, p=0.752021.01 | 65.19 | 69.54 | 83.43 | 71.97 | 72.53 | |
| Deep-all2017.10 | 64.91 | 64.28 | 86.67 | 53.08 | 67.24 | |
| Deep AllBackbone=AlexNet, Reference=from MLDG paper2019.11 | 64.91 | 64.28 | 86.67 | 53.08 | 67.24 | |
| Epi-FCRBackbone=AlexNet, Data Variant=Original2021.01 | 64.7 | 72.3 | 86.1 | 65 | 72.03 | |
| uDICAPipeline=Image -> Deep Feature -> Classifier2017.10 | 64.57 | 64.54 | 91.78 | 51.12 | 68 | |
| Deep AllBackbone=AlexNet, Reference=from D-SAM paper2019.11 | 64.44 | 72.07 | 87.5 | 58.07 | 70.52 | |
| D-SAM*Backbone=AlexNet, Uses domain labels=true2019.11 | 63.87 | 70.7 | 85.55 | 64.66 | 71.2 | |
| Deep-AllPipeline=Image -> E2E2017.10 | 63.3 | 63.13 | 87.7 | 54.07 | 67.05 | |
| Deep AllBackbone=AlexNet2019.11 | 63.3 | 63.13 | 87.7 | 54.07 | 67.05 | |
| Ours-FullPipeline=Image -> E2E2017.10 | 62.86 | 66.97 | 89.5 | 57.51 | 69.21 | |
| AlexNet+TF2017.10 | 62.86 | 66.97 | 89.5 | 57.51 | 69.21 | |
| TF*Backbone=AlexNet, Uses domain labels=true2019.11 | 62.86 | 66.97 | 89.5 | 57.51 | 69.21 | |
| DDAIGBackbone=AlexNet, Data Variant=Original2021.01 | 62.77 | 67.06 | 86.82 | 58.9 | 68.89 | |
| CIDDG*Backbone=AlexNet, Uses domain labels=true2019.11 | 62.7 | 69.73 | 78.65 | 64.45 | 68.88 | |
| Ours-MLPPipeline=Image -> Deep Feature -> Classifier2017.10 | 61.4 | 57.16 | 89.68 | 50.38 | 64.65 | |
| DSNPipeline=Image -> E2E2017.10 | 61.13 | 66.54 | 83.25 | 58.58 | 67.37 | |
| DSN2017.10 | 61.13 | 66.54 | 83.25 | 58.58 | 67.37 | |
| Siamese (RGB)Training source=Cartoon (C)2017.09 | 61 | — | 77 | 51.6 | — | |
| D-MTAE+1HNNPipeline=Image -> Deep Feature -> Classifier2017.10 | 60.27 | 58.65 | 91.12 | 47.86 | 64.48 | |
| D-MTAE2017.10 | 60.27 | 58.65 | 91.12 | 47.86 | 64.48 | |
| LRE-SVMPipeline=Image -> Deep Feature -> Classifier2017.10 | 59.74 | 52.81 | 85.53 | 37.89 | 58.99 | |
| 1HNNPipeline=Image -> Deep Feature -> Classifier2017.10 | 59.1 | 57.89 | 89.86 | 50.31 | 64.29 | |
| Deep AllBackbone=AlexNet, Reference=from CIDDG paper2019.11 | 57.55 | 67.04 | 77.98 | 58.52 | 65.27 | |
| SVMPipeline=Image -> Deep Feature -> Classifier2017.10 | 55.39 | 52.86 | 82.83 | 43.89 | 58.74 | |
| EdgeMACTraining source=All 3 other domains, Input format=edge map2017.09 | 53.8 | 67.9 | 64.5 | 74.7 | 65.2 | |
| EdgeMACTraining source=Photo (P), Input format=edge map2017.09 | 45.4 | 42.3 | — | 46.3 | — | |
| EdgeMACTraining source=Cartoon (C), Input format=edge map2017.09 | 45.2 | — | 57.3 | 74.8 | — | |
| Siamese (RGB)Training source=Sketch (S)2017.09 | 38.7 | 49.3 | 44.4 | — | — | |
| EdgeMACTraining source=Sketch (S), Input format=edge map2017.09 | 34.8 | 63 | 43.3 | — | — | |
| Pre-trained + EdgeMACTraining source=Sketch (S), Fusion=Concatenation2017.09 | 33.7 | 59.3 | 43.4 | — | — | |
| Pre-trained (RGB)Training source=Sketch (S)2017.09 | 31.9 | 49.5 | 42.5 | — | — | |
| EdgeMACTraining source=Art (A), Input format=edge map2017.09 | — | 55.9 | 61.2 | 65.6 | — | |
| Pre-trained (RGB)Training source=Art (A)2017.09 | — | 59.2 | 95 | 33.1 | — | |
| Pre-trained + EdgeMACTraining source=Art (A), Fusion=Concatenation2017.09 | — | 61.6 | 94.9 | 38.4 | — | |
| Siamese (RGB)Training source=Art (A)2017.09 | — | 59.5 | 86.3 | 42.9 | — |