Semantic Image Similarity on CxC
80.4Average Similarity ScoreALIGN-BASE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ALIGN-BASEFine-tuning Data=Multi30K and MSCOCO2021.09 | 80.4 | 0.7 | — | |
| MURAL-LARGEFine-tuning Data=Multi30K and MSCOCO2021.09 | 80.4 | 0.7 | — | |
| MURAL-BASEFine-tuning Data=Multi30K and MSCOCO2021.09 | 80 | 0.7 | — | |
| ALIGN-L2Fine-tuning Data=MSCOCO2021.09 | 77.2 | 0.8 | — | |
| CC:DAEEvaluation Protocol=Zero-shot2024.02 | 74.58 | — | 81.44 | |
| Meaning as TrajectoriesEvaluation Protocol=Zero-shot2024.02 | 74.55 | — | 81.47 | |
| DE-T2T+I2TFine-tuning Data=MSCOCO2021.09 | 74.5 | 0.9 | — | |
| CLIP-ViT B/16Backbone=ViT-B/16, Evaluation Protocol=Zero-shot2024.02 | 68.34 | — | 72.08 | |
| Cond. LikelihoodEvaluation Protocol=Zero-shot2024.02 | — | — | 29.46 |