Image Classification on ImageNet LTNr (IF=100) (val)
81.9Top-1 AccuracyCLIP+LA w. CARE
Evaluation results
| Method | Links | |
|---|---|---|
| CLIP+LA w. CAREImbalance Factor=100, Noise Rate=0.42026.05 | 81.9 | |
| CLIP+CE w. CAREImbalance Factor=100, Noise Rate=0.42026.05 | 81.5 | |
| CLIP+LAImbalance Factor=100, Noise Rate=0.42026.05 | 80.8 | |
| CLIP+CEImbalance Factor=100, Noise Rate=0.42026.05 | 80.5 | |
| CLIP+RLA*Imbalance Factor=100, Noise Rate=0.42026.05 | 79.9 | |
| CLIP zero-shotImbalance Factor=100, Noise Rate=0.42026.05 | 54 | |
| TABASCOImbalance Factor=100, Noise Rate=0.42026.05 | 37.1 | |
| ECBSImbalance Factor=100, Noise Rate=0.42026.05 | 36.9 | |
| ULCImbalance Factor=100, Noise Rate=0.42026.05 | 34.8 | |
| DivdeMixImbalance Factor=100, Noise Rate=0.42026.05 | 34.7 | |
| CEImbalance Factor=100, Noise Rate=0.42026.05 | 31.5 | |
| HARImbalance Factor=100, Noise Rate=0.42026.05 | 31.3 | |
| UNICONImbalance Factor=100, Noise Rate=0.42026.05 | 31.1 | |
| MW-NetImbalance Factor=100, Noise Rate=0.42026.05 | 31.1 | |
| RoLTImbalance Factor=100, Noise Rate=0.42026.05 | 16.9 |