Image Classification on GeoNet target (Asia)
53.9Top-1 Balanced AccuracyCIP+LLM Reg
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CIP+LLM RegEncoder=ViT-B/162024.01 | 53.9 | 1.3 | |
| CIP RegEncoder=ViT-B/162024.01 | 53.5 | 0.9 | |
| LLM RegEncoder=ViT-B/162024.01 | 53.1 | 0.5 | |
| CoCoOpEncoder=ViT-B/162024.01 | 52.6 | — | |
| KgCoOpEncoder=ViT-B/162024.01 | 52.6 | — | |
| CoOpEncoder=ViT-B/162024.01 | 51.2 | — | |
| CIP+LLM RegEncoder=RN502024.01 | 48.3 | 1.4 | |
| CIP RegEncoder=RN502024.01 | 47.6 | 0.7 | |
| LLM RegEncoder=RN502024.01 | 47.4 | 0.5 | |
| KgCoOpEncoder=RN502024.01 | 46.9 | — | |
| CoCoOpEncoder=RN502024.01 | 46.3 | — | |
| CoOpEncoder=RN502024.01 | 45.6 | — |