Text Retrieval on MS COCO 5K
53.7R@1Full-IMRAM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Full-IMRAMvariant=Full model, split=5K2020.03 | 53.7 | 83.2 | 91 | 416.5 | |
| Image-IMRAMvariant=RAM block grounded on images, split=5K2020.03 | 53.2 | 82.5 | 90.4 | 412.7 | |
| VSRN*split=5K2020.03 | 53 | 81.1 | 89.4 | 415.7 | |
| Text-IMRAMvariant=RAM block grounded on texts, split=5K2020.03 | 52 | 81.8 | 90.1 | 409.7 | |
| SCAN*split=5K2020.03 | 50.4 | 82.2 | 90 | 410.9 | |
| PVSEsplit=5K2020.03 | 45.2 | 74.3 | 84.5 | 374.4 | |
| SCOsplit=5K2020.03 | 42.8 | 72.3 | 83 | 369.6 | |
| DPCsplit=5K2020.03 | 41.2 | 70.5 | 81.1 | 337.9 |