Object Detection on COCO Open-vocabulary 2 (test)
45mAP50 (Box, All)Detic
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DeticBackbone=ResNet50-C4, Architecture=Faster R-CNN, IoU threshold=0.5, Supervision=Image labels and captions2022.01 | 45 | 27.8 | 47.1 | |
| Base-onlyBackbone=ResNet50-C4, Architecture=Faster R-CNN, IoU threshold=0.5, Source=Quoted from OVR-CNN [72]2022.01 | 39.9 | 0 | 49.9 | |
| OVR-CNNBackbone=ResNet50-C4, Architecture=Faster R-CNN, IoU threshold=0.5, Source=Quoted from OVR-CNN [72]2022.01 | 39.9 | 22.8 | 46 | |
| Base-only (CLIP)Backbone=ResNet50-C4, Architecture=Faster R-CNN, IoU threshold=0.5, Embeddings=CLIP2022.01 | 39.3 | 1.3 | 48.7 | |
| PLBackbone=ResNet50-C4, Architecture=Faster R-CNN, IoU threshold=0.5, Source=Original publication2022.01 | 27.9 | 4.12 | 35.9 | |
| SBBackbone=ResNet50-C4, Architecture=Faster R-CNN, IoU threshold=0.5, Source=Original publication2022.01 | 24.9 | 0.31 | 29.2 | |
| WSDDNBackbone=ResNet50-C4, Architecture=Faster R-CNN, IoU threshold=0.5, Source=Quoted from OVR-CNN [72]2022.01 | 24.6 | 20.5 | 23.4 | |
| Cap2DetBackbone=ResNet50-C4, Architecture=Faster R-CNN, IoU threshold=0.5, Source=Quoted from OVR-CNN [72]2022.01 | 20.1 | 20.3 | 20.1 | |
| DELOBackbone=ResNet50-C4, Architecture=Faster R-CNN, IoU threshold=0.5, Source=Original publication2022.01 | 13 | 3.41 | 13.8 |