Object Detection on PASCAL VOC (val)
83.1mAP50Partitioned Detector
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Partitioned DetectorTraining Data=COCO, Objects365, OpenImages, Mapillary, Strategy=Partitioned, Backbone=ResNet-50, Architecture=CascadeRCNN2021.02 | 83.1 | — | |
| Unified detector (UniDet)Training Data=COCO, Objects365, OpenImages, Mapillary, Strategy=Unified (retrained), Backbone=ResNet-50, Architecture=CascadeRCNN2021.02 | 82.9 | — | |
| Dataset-specific (Oracle)Training Data=VOC training set, Backbone=ResNet-50, Architecture=CascadeRCNN2021.02 | 80.3 | — | |
| COCO modelTraining Data=COCO, Backbone=ResNet-50, Architecture=CascadeRCNN, Schedule=8x2021.02 | 80 | — | |
| EnsembleTraining Data=COCO, Objects365, OpenImages, Mapillary, Strategy=Ensemble, Backbone=ResNet-50, Architecture=CascadeRCNN2021.02 | 79.7 | — | |
| LP-OVODPre-trained on=LVIS2023.10 | 76 | 59.4 | |
| DetProPre-trained on=LVIS2023.10 | 74.6 | 57.9 | |
| ViLDPre-trained on=LVIS, re-implementation=DetPro repository2023.10 | 72.2 | 56.7 | |
| Objects365 modelTraining Data=Objects365, Backbone=ResNet-50, Architecture=CascadeRCNN, Schedule=8x2021.02 | 71.9 | — | |
| OpenImages modelTraining Data=OpenImages, Backbone=ResNet-50, Architecture=CascadeRCNN, Schedule=8x2021.02 | 64.4 | — | |
| Mapillary modelTraining Data=Mapillary, Backbone=ResNet-50, Architecture=CascadeRCNN, Schedule=8x2021.02 | 11.4 | — |