Image Classification on Caltech-256 (Top-1 and Top-5 Accuracy)
35.49Top-1 AccuracyProxyNCA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ProxyNCALoss function=Proxy Neighborhood Component Analysis2026.03 | 35.49 | — | |
| SupCon2Loss function=Supervised Contrastive (two-view)2026.03 | 34.22 | 54.8 | |
| CE+SIL+SupCon2Loss function=Cross-Entropy + Silhouette + SupCon22026.03 | 33.93 | 52.78 | |
| CELoss function=Cross-Entropy2026.03 | 33.91 | 55.21 | |
| CE+SILLoss function=Cross-Entropy + Silhouette2026.03 | 32.78 | 52.3 | |
| CenterLoss function=Center Loss2026.03 | 32.54 | 53.28 | |
| SupConLoss function=Supervised Contrastive2026.03 | 32.28 | 52.79 |