Classification on ImageNet (test)
55.5kNN AccuracyPiNDA (DINO)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PiNDA (DINO)Backbone=Transformer, PiNDA Setting=N (0, Σ), Batch size=256, Epochs=1002024.08 | 55.5 | 61.22 | |
| PiNDA (SimCLR)Backbone=ResNet18, PiNDA Setting=N (0, Σ), Batch size=256, Epochs=1002024.08 | 55.48 | 66.86 | |
| DINOBackbone=Transformer, PiNDA Setting=—, Batch size=256, Epochs=1002024.08 | 55.44 | 60.65 | |
| SimCLRBackbone=ResNet18, PiNDA Setting=—, Batch size=256, Epochs=1002024.08 | 52.05 | 64.18 | |
| PiNDA (MoCo)Backbone=ResNet18, PiNDA Setting=N (0, Σ), Batch size=256, Epochs=1002024.08 | 51.1 | 64.58 | |
| MoCoBackbone=ResNet18, PiNDA Setting=—, Batch size=256, Epochs=1002024.08 | 50.83 | 64.25 | |
| PiNDA (BYOL)Backbone=ResNet18, PiNDA Setting=N (0, Σ), Batch size=256, Epochs=1002024.08 | 43.25 | 56.57 | |
| PiNDA (SimSiam)Backbone=ResNet18, PiNDA Setting=N (0, Σ), Batch size=256, Epochs=1002024.08 | 43.11 | 56.78 | |
| BYOLBackbone=ResNet18, PiNDA Setting=—, Batch size=256, Epochs=1002024.08 | 43.04 | 56.37 | |
| SimSiamBackbone=ResNet18, PiNDA Setting=—, Batch size=256, Epochs=1002024.08 | 42.96 | 56.55 |