Semantic Similarity on Crisscrossed Captions (CxC)
74.5Mean AverageDE-T2T+I2T
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DE-T2T+I2TFine-tuning Data=MSCOCO2021.09 | 74.5 | — | — | — | 0.4 | |
| MURAL-LARGEFine-tuning Data=Multi30K and MSCOCO2021.09 | 74.1 | — | — | — | 0.4 | |
| MURAL-BASEFine-tuning Data=Multi30K and MSCOCO2021.09 | 73.9 | — | — | — | 0.4 | |
| ALIGN-L2Fine-tuning Data=MSCOCO2021.09 | 72.9 | — | — | — | 0.4 | |
| ALIGN-BASEFine-tuning Data=Multi30K and MSCOCO2021.09 | 72.7 | — | — | — | 0.4 | |
| ALIGN2021.02 | 72.6 | 72.9 | 77.2 | 67.6 | — | |
| DE_T2T+I2T2021.02 | 70.2 | 74.2 | 74.5 | 61.9 | — | |
| VSRN2021.02 | 67.8 | 73 | 70.1 | 60.4 | — | |
| VSE++2021.02 | 67.6 | 74.4 | 73.3 | 55.2 | — | |
| DE_I2T2021.02 | 64.6 | 50.9 | 81.3 | 61.6 | — |