Image Retrieval on WIT translated English (test)
29.98Recall@1GIVL (1M)
Evaluation results
| Method | Links | |
|---|---|---|
| GIVL (1M)Data=3.17M, #Param=112M, Pre-training steps=1M2023.01 | 29.98 | |
| GIVL-BData=3.17M, #Param=112M2023.01 | 29.97 | |
| GIVL w/o LIECData=3.17M, #Param=112M2023.01 | 28.08 | |
| GIVLData=3.17M, #Param=112M2023.01 | 28 | |
| GIVL w/ CLIPData=3.17M, #Param=199M2023.01 | 27.94 | |
| VinVLData=5.65M, #Param=112M2023.01 | 27.78 | |
| GIVL w/o LIKMData=3.17M, #Param=112M2023.01 | 26.21 | |
| VinVL*Data=3.17M, #Param=112M2023.01 | 25.44 | |
| UNITER2023.01 | 15.43 | |
| VILBERTData=3.3M, #Param=274M2023.01 | 15.4 | |
| VisualBERT#Param=135M2023.01 | 15.36 | |
| VL-BERTData=3.3M2023.01 | 15.11 | |
| LXMERT#Param=240M2023.01 | 14.28 |