Image Classification on ImageNet-LTNr (IF=10) (val)
84.6Top-1 AccuracyCLIP+CE w. CARE
Evaluation results
| Method | Links | |
|---|---|---|
| CLIP+CE w. CAREImbalance Factor=10, Noise Rate=0.42026.05 | 84.6 | |
| CLIP+LA w. CAREImbalance Factor=10, Noise Rate=0.42026.05 | 84.4 | |
| CLIP+RLA*Imbalance Factor=10, Noise Rate=0.42026.05 | 84 | |
| CLIP+LAImbalance Factor=10, Noise Rate=0.42026.05 | 83.7 | |
| CLIP+CEImbalance Factor=10, Noise Rate=0.42026.05 | 83.4 | |
| CLIP zero-shotImbalance Factor=10, Noise Rate=0.42026.05 | 54 | |
| ECBSImbalance Factor=10, Noise Rate=0.42026.05 | 50.8 | |
| TABASCOImbalance Factor=10, Noise Rate=0.42026.05 | 49.7 | |
| DivdeMixImbalance Factor=10, Noise Rate=0.42026.05 | 49 | |
| ULCImbalance Factor=10, Noise Rate=0.42026.05 | 47.1 | |
| UNICONImbalance Factor=10, Noise Rate=0.42026.05 | 41.6 | |
| MW-NetImbalance Factor=10, Noise Rate=0.42026.05 | 40.3 | |
| HARImbalance Factor=10, Noise Rate=0.42026.05 | 38.7 | |
| CEImbalance Factor=10, Noise Rate=0.42026.05 | 31.5 | |
| RoLTImbalance Factor=10, Noise Rate=0.42026.05 | 24.2 |