Visual Grounding on RefFLIR RGBT-Ground (test)
72.65Accuracy @ 0.5 IoURGBT-VGNet
Evaluation results
| Method | Links | |
|---|---|---|
| RGBT-VGNetBackbone=CLIP-B / CLIP-B, Modality=RGB+TIR2025.12 | 72.65 | |
| MV-HiVG-LBackbone=CLIP-L / CLIP-L, Modality=RGB+TIR2025.12 | 72.5 | |
| MV-HiVG-BBackbone=CLIP-B / CLIP-B, Modality=RGB+TIR2025.12 | 69.2 | |
| MV-OneRef-BBackbone=BEIT3-B / BEIT3-B, Modality=RGB+TIR2025.12 | 62.05 | |
| MV-OneRef-LBackbone=BEIT3-L / BEIT3-L, Modality=RGB+TIR2025.12 | 60.89 | |
| MV-MMCABackbone=RN50+DETR / BERT-B, Modality=RGB+TIR2025.12 | 48.97 | |
| MV-D-MDETRBackbone=RN50+DETR / BERT-B, Modality=RGB+TIR2025.12 | 47.84 | |
| MV-D-MDETRBackbone=CLIP-B / CLIP-B, Modality=RGB+TIR2025.12 | 47.42 | |
| MV-TransVGBackbone=RN50+DETR / BERT-B, Modality=RGB+TIR2025.12 | 46.06 | |
| MV-CLIP-VGBackbone=CLIP-B / CLIP-B, Modality=RGB+TIR2025.12 | 46.01 |