Caption Matching on XTD-10 v1 (test)
41.8Score (de)CLIP
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CLIPMode=Zero-shot2024.01 | 41.8 | 81.5 | 50.2 | 49.4 | 41 | 13.2 | 0.5 | 5.4 | 0.8 | 4.3 | 2.7 | 26.4 | |
| QAP MatchingMethod Variation=QAP, Evaluation Protocol=Training-free transfer2024.01 | 39.6 | 51.6 | 41.4 | 40.2 | 38.5 | 30.5 | 30.9 | 40.2 | 35.1 | 38.9 | 40.3 | 38.8 | |
| RelativeCitation=[34]2024.01 | 35 | 52.5 | 37.8 | 37.5 | 37.2 | 28.3 | 30.4 | 36.6 | 31.9 | 35.8 | 36.5 | 36.3 | |
| LinearEvaluation Protocol=Linear Probing2024.01 | 34 | 40.9 | 31.7 | 30.7 | 34.9 | 23.5 | 23.5 | 30.2 | 30.7 | 29.6 | 31.1 | 30.9 |