KNN Classification on CIFAR-100 (test)
77.7Top-1 AccuracyGrafit
Evaluation results
| Method | Links | |
|---|---|---|
| GrafitBackbone=ResNet-50, Number of parameters=32.9M2020.11 | 77.7 | |
| Grafit FCBackbone=ResNet-50, Number of parameters=~23.5M2020.11 | 75.6 | |
| ClusterFit+Backbone=ResNet-50, Number of parameters=32.9M2020.11 | 72.5 | |
| SNCA+Backbone=ResNet-50, Number of parameters=32.9M2020.11 | 72.2 | |
| Baseline (ours)Backbone=ResNet-50, Number of parameters=~23.5M2020.11 | 71.8 | |
| SimSiamTraining Protocol=Cut (c = 9), Epochs=10002025.02 | 62.6 | |
| SNCA, Wu et al.Backbone=ResNet-50, Number of parameters=~23.5M2020.11 | 62.3 | |
| SimSiamTraining Protocol=Default, Epochs=10002025.02 | 61.8 | |
| SimCLRTraining Protocol=Cut (c = 3), Epochs=10002025.02 | 60.1 | |
| SimCLRTraining Protocol=GradScale, Epochs=10002025.02 | 58.2 | |
| SimCLRTraining Protocol=Default, Epochs=10002025.02 | 56.6 | |
| Baseline, Wu et al.Backbone=ResNet-50, Number of parameters=~23.5M2020.11 | 54.2 |