Object Detection on Cityscapes -> BDD100K (val)
58.003AP@0.5Full-Target Oracle
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Full-Target OracleDetector=DINO, Target-label budget=5%2026.06 | 58.003 | — | — | — | — | — | — | — | — | — | — | |
| RTSM + Multi-View Foreground Revival (DDT)SFDA-OD method=DDT, Detector=DINO, Target-label budget=5%2026.06 | 50.462 | — | — | — | — | — | — | — | — | — | 2.049 | |
| RTSM + Target-Anchored PCGrad (DDT)SFDA-OD method=DDT, Detector=DINO, Target-label budget=5%2026.06 | 50.252 | — | — | — | — | — | — | — | — | — | 1.84 | |
| RTSM + Label-Guided Teacher Update (DDT)SFDA-OD method=DDT, Detector=DINO, Target-label budget=5%2026.06 | 49.788 | — | — | — | — | — | — | — | — | — | 1.375 | |
| RTSM + Sparse Loss Balancing (DDT)SFDA-OD method=DDT, Detector=DINO, Target-label budget=5%2026.06 | 49.213 | — | — | — | — | — | — | — | — | — | 0.8 | |
| RTSM + Prior-Mapped Thresholding (DDT)SFDA-OD method=DDT, Detector=DINO, Target-label budget=5%2026.06 | 49.195 | — | — | — | — | — | — | — | — | — | 0.783 | |
| RTSM + Foreground Query Revival (DDT)SFDA-OD method=DDT, Detector=DINO, Target-label budget=5%2026.06 | 48.526 | — | — | — | — | — | — | — | — | — | 0.114 | |
| RTSM (DDT)SFDA-OD method=DDT, Detector=DINO, Target-label budget=5%2026.06 | 48.413 | — | — | — | — | — | — | — | — | — | — | |
| RTSM + Hard Query Recovery (DDT)SFDA-OD method=DDT, Detector=DINO, Target-label budget=5%2026.06 | 46.628 | — | — | — | — | — | — | — | — | — | -1.784 | |
| RTSM + Multi-View Hard Recovery (DDT)SFDA-OD method=DDT, Detector=DINO, Target-label budget=5%2026.06 | 45.834 | — | — | — | — | — | — | — | — | — | -2.578 | |
| RTSM + Precision-Calibrated Thresholding (DDT)SFDA-OD method=DDT, Detector=DINO, Target-label budget=5%2026.06 | 44.814 | — | — | — | — | — | — | — | — | — | -3.598 | |
| RTSM + Reliability Score Rescaling (DDT)SFDA-OD method=DDT, Detector=DINO, Target-label budget=5%2026.06 | 44.015 | — | — | — | — | — | — | — | — | — | -4.398 | |
| RTSM + Multi-View Hard Recovery (LPU)SFDA-OD method=LPU, Detector=DINO, Target-label budget=5%2026.06 | 41.578 | — | — | — | — | — | — | — | — | — | 2.221 | |
| RTSM + Hard Query Recovery (PETS)SFDA-OD method=PETS, Detector=DINO, Target-label budget=5%2026.06 | 41.164 | — | — | — | — | — | — | — | — | — | 0.631 | |
| RTSM + Hard Query Recovery (LPU)SFDA-OD method=LPU, Detector=DINO, Target-label budget=5%2026.06 | 41 | — | — | — | — | — | — | — | — | — | 1.644 | |
| RTSM + Prior-Mapped Thresholding (LPU)SFDA-OD method=LPU, Detector=DINO, Target-label budget=5%2026.06 | 40.988 | — | — | — | — | — | — | — | — | — | 1.632 | |
| RTSM (PETS)SFDA-OD method=PETS, Detector=DINO, Target-label budget=5%2026.06 | 40.533 | — | — | — | — | — | — | — | — | — | — | |
| RTSM + Sparse Loss Balancing (PETS)SFDA-OD method=PETS, Detector=DINO, Target-label budget=5%2026.06 | 40.324 | — | — | — | — | — | — | — | — | — | -0.209 | |
| RTSM + Precision-Calibrated Thresholding (LPU)SFDA-OD method=LPU, Detector=DINO, Target-label budget=5%2026.06 | 40.305 | — | — | — | — | — | — | — | — | — | 0.949 | |
| RTSM + Target-Anchored PCGrad (PETS)SFDA-OD method=PETS, Detector=DINO, Target-label budget=5%2026.06 | 40.287 | — | — | — | — | — | — | — | — | — | -0.246 | |
| RTSM + Prior-Mapped Thresholding (PETS)SFDA-OD method=PETS, Detector=DINO, Target-label budget=5%2026.06 | 40.227 | — | — | — | — | — | — | — | — | — | -0.306 | |
| RTSM + Label-Guided Teacher Update (PETS)SFDA-OD method=PETS, Detector=DINO, Target-label budget=5%2026.06 | 40.14 | — | — | — | — | — | — | — | — | — | -0.394 | |
| RTSM + Reliability Score Rescaling (PETS)SFDA-OD method=PETS, Detector=DINO, Target-label budget=5%2026.06 | 40.136 | — | — | — | — | — | — | — | — | — | -0.397 | |
| RTSM + Hard Query Recovery (LPLD)SFDA-OD method=LPLD, Detector=DINO, Target-label budget=5%2026.06 | 40.128 | — | — | — | — | — | — | — | — | — | 2.072 | |
| RTSM + Precision-Calibrated Thresholding (PETS)SFDA-OD method=PETS, Detector=DINO, Target-label budget=5%2026.06 | 40.053 | — | — | — | — | — | — | — | — | — | -0.48 | |
| RTSM + Multi-View Foreground Revival (PETS)SFDA-OD method=PETS, Detector=DINO, Target-label budget=5%2026.06 | 39.828 | — | — | — | — | — | — | — | — | — | -0.705 | |
| RTSM + Target-Anchored PCGrad (LPU)SFDA-OD method=LPU, Detector=DINO, Target-label budget=5%2026.06 | 39.743 | — | — | — | — | — | — | — | — | — | 0.387 | |
| RTSM + Foreground Query Revival (PETS)SFDA-OD method=PETS, Detector=DINO, Target-label budget=5%2026.06 | 39.614 | — | — | — | — | — | — | — | — | — | -0.919 | |
| RTSM (LPU)SFDA-OD method=LPU, Detector=DINO, Target-label budget=5%2026.06 | 39.356 | — | — | — | — | — | — | — | — | — | — | |
| RTSM + Multi-View Foreground Revival (LPU)SFDA-OD method=LPU, Detector=DINO, Target-label budget=5%2026.06 | 39.232 | — | — | — | — | — | — | — | — | — | -0.125 | |
| RTSM + Precision-Calibrated Thresholding (LPLD)SFDA-OD method=LPLD, Detector=DINO, Target-label budget=5%2026.06 | 39.064 | — | — | — | — | — | — | — | — | — | 1.009 | |
| RTSM + Target-Anchored PCGrad (LPLD)SFDA-OD method=LPLD, Detector=DINO, Target-label budget=5%2026.06 | 38.91 | — | — | — | — | — | — | — | — | — | 0.855 | |
| RTSM + Prior-Mapped Thresholding (LPLD)SFDA-OD method=LPLD, Detector=DINO, Target-label budget=5%2026.06 | 38.894 | — | — | — | — | — | — | — | — | — | 0.839 | |
| RTSM + Sparse Loss Balancing (LPU)SFDA-OD method=LPU, Detector=DINO, Target-label budget=5%2026.06 | 38.724 | — | — | — | — | — | — | — | — | — | -0.632 | |
| RTSM + Label-Guided Teacher Update (LPU)SFDA-OD method=LPU, Detector=DINO, Target-label budget=5%2026.06 | 38.679 | — | — | — | — | — | — | — | — | — | -0.677 | |
| RTSM + Reliability Score Rescaling (LPU)SFDA-OD method=LPU, Detector=DINO, Target-label budget=5%2026.06 | 38.585 | — | — | — | — | — | — | — | — | — | -0.772 | |
| RTSM + Foreground Query Revival (LPU)SFDA-OD method=LPU, Detector=DINO, Target-label budget=5%2026.06 | 38.403 | — | — | — | — | — | — | — | — | — | -0.953 | |
| RTSM + Multi-View Foreground Revival (LPLD)SFDA-OD method=LPLD, Detector=DINO, Target-label budget=5%2026.06 | 38.068 | — | — | — | — | — | — | — | — | — | 0.013 | |
| RTSM (LPLD)SFDA-OD method=LPLD, Detector=DINO, Target-label budget=5%2026.06 | 38.055 | — | — | — | — | — | — | — | — | — | — | |
| RTSM + Label-Guided Teacher Update (LPLD)SFDA-OD method=LPLD, Detector=DINO, Target-label budget=5%2026.06 | 38 | — | — | — | — | — | — | — | — | — | -0.055 | |
| RTSM + Reliability Score Rescaling (LPLD)SFDA-OD method=LPLD, Detector=DINO, Target-label budget=5%2026.06 | 37.98 | — | — | — | — | — | — | — | — | — | -0.075 | |
| RTSM + Multi-View Hard Recovery (LPLD)SFDA-OD method=LPLD, Detector=DINO, Target-label budget=5%2026.06 | 37.792 | — | — | — | — | — | — | — | — | — | -0.263 | |
| RTSM + Foreground Query Revival (LPLD)SFDA-OD method=LPLD, Detector=DINO, Target-label budget=5%2026.06 | 37.637 | — | — | — | — | — | — | — | — | — | -0.418 | |
| RTSM + Sparse Loss Balancing (LPLD)SFDA-OD method=LPLD, Detector=DINO, Target-label budget=5%2026.06 | 36.678 | — | — | — | — | — | — | — | — | — | -1.378 | |
| PETSProtocol=Pure SFDA, Detector=DINO, Target-label budget=5%2026.06 | 36.153 | — | — | — | — | — | — | — | — | — | — | |
| DDTProtocol=Pure SFDA, Detector=DINO, Target-label budget=5%2026.06 | 34.77 | — | — | — | — | — | — | — | — | — | — | |
| LPLDProtocol=Pure SFDA, Detector=DINO, Target-label budget=5%2026.06 | 28.465 | — | — | — | — | — | — | — | — | — | — | |
| RTSM + Multi-View Hard Recovery (PETS)SFDA-OD method=PETS, Detector=DINO, Target-label budget=5%2026.06 | 26.974 | — | — | — | — | — | — | — | — | — | -13.56 | |
| Source-OnlyDetector=DINO, Target-label budget=5%2026.06 | 26.225 | — | — | — | — | — | — | — | — | — | — | |
| LPUProtocol=Pure SFDA, Detector=DINO, Target-label budget=5%2026.06 | 21.067 | — | — | — | — | — | — | — | — | — | — | |
| CR-DABackbone=VGG-162022.06 | — | 19.5 | 46.3 | 31.3 | 31.4 | 17.3 | 23.8 | 18.9 | — | 26.9 | — | |
| CR-DADetector=Faster RCNN2022.04 | — | 30.8 | 29 | 44.8 | 20.5 | 19.8 | — | 14.1 | 22.8 | 26 | — | |
| CR-SWDetector=Faster RCNN2022.04 | — | 32.8 | 29.3 | 45.8 | 22.7 | 20.6 | — | 14.9 | 25.5 | 27.4 | — | |
| DAFDetector=Faster RCNN2022.04 | — | 28.9 | 27.4 | 44.2 | 19.1 | 18 | — | 14.2 | 22.4 | 24.9 | — | |
| Deformable DETR (source)Detector=D-DETR2022.04 | — | 38.9 | 26.7 | 55.2 | 15.7 | 19.7 | — | 10.8 | 16.2 | 26.2 | — | |
| EPMDetector=FCOS2022.04 | — | 39.6 | 26.8 | 55.8 | 18.8 | 19.1 | — | 14.5 | 20.1 | 27.8 | — | |
| Faster R-CNN (source)Detector=Faster RCNN2022.04 | — | 28.8 | 25.4 | 44.1 | 17.9 | 16.1 | — | 13.9 | 22.4 | 24.1 | — | |
| FCOS (source)Detector=FCOS2022.04 | — | 38.6 | 24.8 | 54.5 | 17.2 | 16.3 | — | 15 | 18.3 | 26.4 | — | |
| LabelMatchBackbone=VGG-16, Class distribution estimation=labeled source data2022.06 | — | 39.4 | 54.6 | 37.4 | 42.9 | 25.7 | 29.8 | 41.7 | — | 38.8 | — | |
| LabelMatch†Backbone=VGG-16, Class distribution estimation=GT labels of unlabeled data (ideal setting)2022.06 | — | 39.8 | 55.4 | 44.5 | 44.8 | 38.6 | 41.5 | 47.1 | — | 44.5 | — | |
| O2netDetector=D-DETR2022.04 | — | 40.4 | 31.2 | 58.6 | 20.4 | 25 | — | 14.9 | 22.7 | 30.5 | — | |
| SCDADetector=Faster RCNN2022.04 | — | 29.3 | 29.2 | 44.4 | 20.3 | 19.6 | — | 14.8 | 23.2 | 25.8 | — | |
| SFADetector=D-DETR2022.04 | — | 40.2 | 27.6 | 57.5 | 19.1 | 23.4 | — | 15.4 | 19.2 | 28.9 | — | |
| Source onlyBackbone=VGG-162022.06 | — | 14.7 | 50 | 33.3 | 35.8 | 18.4 | 27.6 | 17 | — | 28.7 | — | |
| SW-FasterBackbone=VGG-162022.06 | — | 15.2 | 45.7 | 29.5 | 30.2 | 17.1 | 21.2 | 18.4 | — | 25.3 | — | |
| SWDADetector=Faster RCNN2022.04 | — | 29.5 | 29.9 | 44.8 | 20.2 | 20.7 | — | 15.2 | 23.1 | 26.2 | — |