Image Classification on CIFAR-100 long-tailed (test) (Recall Metrics)
0.489Average RecallLA with class priors
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LA with class priorsBackbone=ResNet-56, Objective=Maximizing worst-case recall2021.07 | 0.489 | 0.2 | |
| CSL [Logit-adjusted]Backbone=ResNet-56, Objective=Maximizing worst-case recall2021.07 | 0.466 | 0.418 | |
| ERMBackbone=ResNet-56, Objective=Maximizing worst-case recall2021.07 | 0.444 | 0.076 | |
| CSL [Re-weighted]Backbone=ResNet-56, Objective=Maximizing worst-case recall2021.07 | 0.415 | 0.083 |