Image Classification on Cross-Domain Classification Suite (12 datasets) (test)
49.89ChestX AccuracySL-MLP
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SL-MLPEvaluation Protocol=Linear evaluation, Backbone=Fixed, Pre-training Epochs=3002021.12 | 49.89 | 99.02 | 87.86 | 72.61 | 96.63 | 93.46 | 81.12 | 76.73 | 94.2 | 85.36 | 67.82 | 67.13 | 81.15 | |
| SelfSupCon†Evaluation Protocol=Linear evaluation, Backbone=Fixed, Pre-training Epochs=4002021.12 | 48.08 | 99.06 | 87.88 | 72.71 | 96.97 | 89.62 | 81.67 | 69.66 | 90.88 | 69.12 | 69.95 | 81.51 | 79.7 | |
| SupConEvaluation Protocol=Linear evaluation, Backbone=Fixed, Pre-training Epochs=3002021.12 | 47.71 | 98.79 | 85.66 | 74.2 | 95.83 | 92.24 | 79.42 | 73.42 | 91.14 | 76.8 | 74.26 | 79.78 | 80.77 | |
| SLEvaluation Protocol=Linear evaluation, Backbone=Fixed, Pre-training Epochs=3002021.12 | 45.45 | 96.8 | 84.02 | 66.22 | 95.07 | 83.69 | 75.4 | 64.14 | 91.66 | 74.51 | 75.16 | 81.53 | 81.68 | |
| SelfSupCon†Evaluation Protocol=Finetuned with 1000 training samples, Backbone=Full network, Pre-training Epochs=4002021.12 | 43.09 | 93.95 | 88.1 | 62.95 | 95.47 | 88.92 | 79.41 | 45.33 | 81.14 | 10.57 | 82.37 | 78.27 | 70.88 | |
| SL-MLPEvaluation Protocol=Finetuned with 1000 training samples, Backbone=Full network, Pre-training Epochs=3002021.12 | 42.34 | 94.48 | 89.64 | 63.9 | 95.3 | 90.2 | 77.98 | 46.66 | 83.13 | 17.32 | 80.19 | 78.82 | 71.66 | |
| SupConEvaluation Protocol=Finetuned with 1000 training samples, Backbone=Full network, Pre-training Epochs=3002021.12 | 41.84 | 93.46 | 88.7 | 61.81 | 94.54 | 91.28 | 78.35 | 46.02 | 81.62 | 15.84 | 81.85 | 78.51 | 71.15 | |
| SupCon w/o MLPEvaluation Protocol=Finetuned with 1000 training samples, Backbone=Full network, Pre-training Epochs=3002021.12 | 41.72 | 93.52 | 84.95 | 58.09 | 95.15 | 88.23 | 78.95 | 45.68 | 80.63 | 14.39 | 82.25 | 77.96 | 70.12 | |
| SupCon w/o MLPEvaluation Protocol=Linear evaluation, Backbone=Fixed, Pre-training Epochs=3002021.12 | 41.38 | 91.52 | 73.16 | 62.93 | 89.84 | 73.23 | 66.38 | 44.54 | 76.55 | 55.21 | 61.45 | 68.54 | 67.06 | |
| SLEvaluation Protocol=Finetuned with 1000 training samples, Backbone=Full network, Pre-training Epochs=3002021.12 | 40.86 | 94.31 | 86.95 | 62.12 | 94.05 | 88.94 | 78.22 | 46.16 | 80.32 | 14.17 | 82.16 | 78.28 | 70.54 | |
| SL-MLPEvaluation Protocol=5-ways 5-shots few-shot classification, Pre-training Epochs=3002021.12 | 26.89 | 93.45 | 59.08 | 83.04 | 87.16 | 96.88 | 50.77 | 95.73 | 89 | 89.84 | 41.96 | 46.76 | 71.71 | |
| SupConEvaluation Protocol=5-ways 5-shots few-shot classification, Pre-training Epochs=3002021.12 | 26.18 | 94.09 | 59.36 | 85.02 | 87.97 | 96.55 | 51.02 | 94.49 | 89.01 | 89.75 | 41.67 | 43.48 | 71.55 | |
| SLEvaluation Protocol=5-ways 5-shots few-shot classification, Pre-training Epochs=3002021.12 | 25.64 | 89.07 | 54.32 | 78.58 | 82.96 | 93.14 | 46.14 | 92.82 | 84.17 | 87.06 | 38.03 | 41.22 | 67.76 | |
| SupCon w/o MLPEvaluation Protocol=5-ways 5-shots few-shot classification, Pre-training Epochs=3002021.12 | 23.62 | 75.64 | 49.34 | 73.04 | 73.9 | 82.16 | 38.1 | 67.87 | 75.18 | 81.01 | 34.92 | 35.16 | 59.16 |