Image Classification on 11 Transfer Classification Datasets (test)
83.1Accuracy (Aircraft)SwAV
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SwAVBackbone=ResNet, Evaluation Protocol=Fine-tuned2023.03 | 83.1 | 89.9 | 86.8 | 96.8 | 84.4 | 75.2 | 95.5 | 87.2 | 89.1 | 66.2 | 84.7 | 85.3 | — | |
| SDRBackbone=ResNet, Evaluation Protocol=Fine-tuned2023.03 | 82.6 | 89 | 87.5 | 97.4 | 84.4 | 75.6 | 97 | 86.1 | 89.3 | 66.1 | 85.3 | 85.5 | — | |
| MixedAEBackbone=Transformer, Evaluation Protocol=Fine-tuned, Pre-training Epochs=16002023.03 | 82.1 | 91.5 | 88.8 | 97.9 | 85.9 | 78.7 | 97.1 | 87.4 | 93.6 | 66.2 | 86.4 | 86.9 | — | |
| MoCov2Backbone=ResNet, Evaluation Protocol=Fine-tuned2023.03 | 79.9 | 84.4 | 75.2 | 96.5 | 71.3 | 69.5 | 94.4 | 76.8 | 79.8 | 55.8 | 71.7 | 77.7 | — | |
| BYOLBackbone=ResNet, Evaluation Protocol=Fine-tuned2023.03 | 79.5 | 89.4 | 84.6 | 97 | 84 | 73.6 | 94.5 | 85.5 | 89.6 | 64 | 82.7 | 84 | — | |
| SimCLRBackbone=ResNet, Evaluation Protocol=Fine-tuned2023.03 | 78.7 | 82.9 | 79.8 | 96.2 | 79.1 | 70.2 | 94.3 | 82.2 | 83.2 | 61.1 | 78.2 | 80.5 | — | |
| MAEBackbone=Transformer, Evaluation Protocol=Fine-tuned2023.03 | 78.2 | 91.2 | 88.4 | 97 | 82.5 | 75.3 | 96.6 | 84.7 | 92.6 | 65.4 | 86 | 85.3 | — | |
| MoCov3Backbone=Transformer, Evaluation Protocol=Fine-tuned2023.03 | 76.6 | 91.2 | 86.6 | 98.3 | 88.3 | 72.6 | 95.5 | 86.4 | 92 | 65.6 | 84.5 | 85.2 | — | |
| DINOBackbone=Transformer, Evaluation Protocol=Fine-tuned2023.03 | 69.4 | 91.2 | 81.3 | 98.4 | 88.9 | 77.6 | 96.9 | 87.3 | 93.5 | 64.7 | 86.3 | 85.1 | — | |
| BEITBackbone=Transformer, Evaluation Protocol=Fine-tuned2023.03 | 66.3 | 80.2 | 78.6 | 96.1 | 80 | 69.9 | 92.9 | 83.2 | 85.3 | 57.1 | 76.7 | 78.7 | — | |
| GroCoEpochs=400, Batch Size=1024, Backbone=ResNet-50, Views=2x224, Evaluation Protocol=Zero-shot k-NN2023.01 | 27.6 | 81.2 | 19.1 | 86.8 | 65.8 | 71 | 79.2 | 56.3 | 80.6 | 52.3 | 74.6 | 63.14 | — | |
| SwAVEpochs=400, Batch Size=4096, Backbone=ResNet-50, Views=2x224, Evaluation Protocol=Zero-shot k-NN2023.01 | 25.8 | 82.2 | 17.8 | 88.6 | 66 | 71.1 | 74.8 | 50.3 | 70.4 | 52.9 | 75.6 | 61.41 | — | |
| SimSiamEpochs=100, Batch Size=256, Backbone=ResNet-50, Views=2x224, Evaluation Protocol=Zero-shot k-NN2023.01 | 25.4 | 80.1 | 17.3 | 87.4 | 65.5 | 69.3 | 77.3 | 52.5 | 72.5 | 50 | 72.8 | 60.92 | — | |
| MoCo v2Epochs=200, Batch Size=256, Backbone=ResNet-50, Views=2x224, Evaluation Protocol=Zero-shot k-NN2023.01 | 20.1 | 78.9 | 12.3 | 86.9 | 61.4 | 68.7 | 67.3 | 47.2 | 67.2 | 46.9 | 73.9 | 57.35 | — | |
| SimCLREpochs=400, Batch Size=4096, Backbone=ResNet-50, Views=2x224, Evaluation Protocol=Zero-shot k-NN2023.01 | 19.3 | 77.7 | 15.3 | 85.4 | 61 | 70.2 | 72.5 | 51.2 | 68.2 | 49 | 73.1 | 58.45 | — |