Phrase Grounding on ReferIt (test)
80.58Pointing AccuracyBARE-L
Evaluation Results
| Method | Links | |
|---|---|---|
| BARE-LVisual Encoder=BEiT-L [9], Fine-tuning protocol=Fine-tuning, Pre-training type=pretrained open-set detection model, Training dataset scope=mixed datasets2026.01 | 80.58 | |
| WWbLBackbone=CLIP+VGG, Training source=VG2022.06 | 65.95 | |
| MGBackbone=BiLSTM+VGG, Training source=VG2022.06 | 62.76 | |
| WWbLBackbone=CLIP+VGG, Training source=MS COCO2022.06 | 61.03 | |
| MGBackbone=ELMo+VGG, Training source=VG2022.06 | 60.01 | |
| GbSBackbone=VGG, Training source=VG2022.06 | 59.44 | |
| GAEBackbone=CLIP, Training source=VG2022.06 | 56.76 | |
| GAEBackbone=CLIP, Training source=MS COCO2022.06 | 56.76 | |
| GbSBackbone=VGG, Training source=MS COCO2022.06 | 56.1 | |
| MGBackbone=BiLSTM+VGG, Training source=MS COCO2022.06 | 47.89 | |
| MGBackbone=ELMo+VGG, Training source=MS COCO2022.06 | 47.52 | |
| SSSBackbone=VGG, Training source=VG2022.06 | 39.98 | |
| FCVCBackbone=VGG, Training source=MS COCO2022.06 | 33.52 | |
| TDBackbone=Inception-2, Training source=VG2022.06 | 31.97 | |
| Baseline CenterBackbone=Center, Training source=VG2022.06 | 30.3 | |
| Baseline CenterBackbone=Center, Training source=MS COCO2022.06 | 30.3 | |
| Baseline RandomBackbone=Random, Training source=VG2022.06 | 24.3 | |
| Baseline RandomBackbone=Random, Training source=MS COCO2022.06 | 24.3 |