Image Classification on ImageNet 1K (val) (Top-1 Accuracy at 100 & 200 eps)
69.7Top-1 Acc (100 eps)CW-RGP 4
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CW-RGP 4Batch size=512, Encoder=ResNet-50, Evaluation Protocol=linear classification2022.10 | 69.7 | 71 | |
| W-MSE 4Batch size=4096, Encoder=ResNet-50, Evaluation Protocol=linear classification2022.10 | 69.4 | — | |
| Zero-CLBatch size=1024, Encoder=ResNet-50, Evaluation Protocol=linear classification2022.10 | 68.9 | — | |
| SimSiamBatch size=256, Encoder=ResNet-50, Evaluation Protocol=linear classification2022.10 | 68.1 | 70 | |
| MoCo v2Batch size=256, Encoder=ResNet-50, Evaluation Protocol=linear classification2022.10 | 67.4 | 69.9 | |
| W-MSE 4 (repro.)Batch size=512, Encoder=ResNet-50, Evaluation Protocol=linear classification2022.10 | 66.7 | 67.9 | |
| SimCLRBatch size=4096, Encoder=ResNet-50, Evaluation Protocol=linear classification2022.10 | 66.5 | 68.3 | |
| BYOLBatch size=4096, Encoder=ResNet-50, Evaluation Protocol=linear classification2022.10 | 66.5 | 70.6 | |
| SwAVBatch size=4096, Encoder=ResNet-50, Evaluation Protocol=linear classification2022.10 | 66.5 | 69.1 | |
| BYOL (repro.)Batch size=512, Encoder=ResNet-50, Evaluation Protocol=linear classification2022.10 | 66.1 | 69.2 | |
| SwAV (repro.)Batch size=512, Encoder=ResNet-50, Evaluation Protocol=linear classification2022.10 | 65.8 | 67.9 |