Object Detection on Object365
39.3APFlorence
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| FlorenceInference Mode=Fine-tuning2021.11 | 39.3 | — | — | — | |
| Multi-dataset DetectionInference Mode=Fine-tuning2021.11 | 33.7 | — | — | — | |
| IntRecTurn=12026.02 | 17.2 | 23.9 | — | 14.7 | |
| DECOLA Phase 2Backbone=ResNet-50, Architecture=DETR, Training Data=LVIS-base and weakly-labeled data2023.11 | 15 | 22 | 16 | — | |
| DetLH2026.02 | 14.6 | 21.2 | — | 11.9 | |
| DeticBackbone=ResNet-50, Architecture=DETR, Training Data=LVIS-base and weakly-labeled data2023.11 | 14.5 | 21.4 | 15.5 | — | |
| DeticBackbone=ResNet-50, Architecture=R-CNN, Training Data=LVIS-base and weakly-labeled data2023.11 | 14.2 | 20.7 | 15.2 | — | |
| CoDet2026.02 | 13.8 | 20.6 | — | 10.5 | |
| IntRecTurn=02026.02 | 13.8 | 21.4 | — | 11.5 | |
| BARONBackbone=ResNet-50, Architecture=R-CNN, Training Data=LVIS-base and weakly-labeled data2023.11 | 13.6 | 21 | 14.5 | — | |
| CCKT-Det2026.02 | 13.4 | 19.7 | — | 11.5 | |
| BARON2026.02 | 12.7 | 20.3 | — | 10.2 | |
| OVMR2026.02 | 12.3 | 19.5 | — | 10.7 | |
| DetProBackbone=ResNet-50, Architecture=R-CNN, Training Data=LVIS-base and weakly-labeled data2023.11 | 12.1 | 18.8 | 12.9 | — | |
| F-VLMBackbone=ResNet-50, Architecture=R-CNN, Training Data=LVIS-base and weakly-labeled data2023.11 | 11.9 | 19.2 | 12.6 | — | |
| VILDBackbone=ResNet-50, Architecture=R-CNN, Training Data=LVIS-base and weakly-labeled data2023.11 | 11.8 | 18.2 | 12.6 | — | |
| DetPro2026.02 | 11.7 | 18.2 | — | 9.8 | |
| Gao et al.Iterations x Batch size=150K×64, Open-vocabulary=true2022.07 | — | 6.9 | — | — | |
| VL-PLMIterations x Batch size=180K×16, Open-vocabulary=true2022.07 | — | 10.9 | — | — |