Object Detection on COCO zero-shot 2017
34.4Novel APVL-PLM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| VL-PLMTraining Source=raw image-text pairs via Internet, instance-level labels in SB2022.07 | 34.4 | 60.2 | 53.5 | |
| RegionCLIPTraining Source=image-caption pairs in SB ∪ SN, instance-level labels in SB2022.07 | 31.4 | 57.1 | 50.4 | |
| Gao et al.Training Source=raw image-text pairs via Internet2022.07 | 30.8 | 46.1 | 42.1 | |
| ViLDTraining Source=raw image-text pairs via Internet, instance-level labels in SB2022.07 | 27.6 | 59.5 | 51.3 | |
| OVR-CNNTraining Source=image-caption pairs in SB ∪ SN, instance-level labels in SB2022.07 | 22.8 | 46 | 39.9 | |
| RegionCLIP*Refinement using image-caption pairs=false2022.07 | 14.2 | 52.8 | 42.7 | |
| Rahman et al.Training Source=instance-level labels in SB2022.07 | 4.12 | 35.9 | 27.9 | |
| Zhu et al.Training Source=instance-level labels in SB2022.07 | 3.41 | 13.8 | 13 | |
| Bansal et al.2022.07 | 0.31 | 29.2 | 24.9 |