Image Classification on IMAGENET-LT ResNet-50 (test)
62Top-1 AccuracyREPAIR
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| REPAIRBackbone=ResNet-50, Shortlist size=102026.04 | 62 | 85.4 | 75.1 | 60.8 | 65.7 | 37.8 | |
| ClasswiseBackbone=ResNet-50, Shortlist size=102026.04 | 61.8 | 85.4 | 75 | 59.9 | 67.8 | 33.6 | |
| LogitAdjBackbone=ResNet-50, Shortlist size=102026.04 | 61.5 | 85.3 | 74.8 | 59.5 | 68.1 | 32.1 | |
| τ-normBackbone=ResNet-50, Shortlist size=102026.04 | 60.5 | 84.8 | 74.1 | 58.1 | 68.2 | 29.9 | |
| BaseBackbone=ResNet-50, Shortlist size=102026.04 | 54.4 | 78.6 | 68.9 | 47.4 | 76.7 | — |