Open-vocabulary Object Detection on COCO OVD (48 base 17 novel)
36.3mAP (Novel)RegionCLIP + APT(v_i)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RegionCLIP + APT(v_i)Backbone=ResNet50, Variant=v_i2023.05 | 36.3 | 57.3 | |
| RegionCLIP + APT(v_i + o_i)Backbone=ResNet50, Variant=v_i + o_i2023.05 | 35.9 | 57.7 | |
| RegionCLIPBackbone=ResNet502023.05 | 35.2 | 57.6 |