Referring Expression Comprehension on RefCOCO (test)
81.2AccuracyOur SAT->ST
Evaluation results
| Method | Links | |
|---|---|---|
| Our SAT->STBackbone=BERTB, Training Strategy=All-Task Pretraining + Single-Task Finetuning, Number of Parameters=3B, Number of Models=12 x 250M2019.12 | 81.2 | |
| UNITER [8]Backbone=BERTL, Number of Parameters=2.1B, Number of Models=7 x 303M2019.12 | 80.88 | |
| Our SATBackbone=BERTB, Training Strategy=All-Task (AT), Number of Parameters=270M, Number of Models=1 x 270M2019.12 | 80.58 | |
| UNITER [8]Backbone=BERTB, Number of Parameters=602M, Number of Models=7 x 86M2019.12 | 80.21 |