Image Classification on ImageNet (test) (Accuracy and GPU Savings)
75.92Top-1 AccKnee schedule
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Knee scheduleBudget=50 epochs, Backbone=ResNet-502020.03 | 75.92 | 27 | |
| Linear DecayBudget=50 epochs, Backbone=ResNet-502020.03 | 75.82 | — | |
| Cosine DecayBudget=50 epochs, Backbone=ResNet-502020.03 | 75.71 | — | |
| One-CycleBudget=50 epochs, Backbone=ResNet-502020.03 | 75.36 | — |