Object Detection on NuImages low-performing categories
40mAPOurs
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| OursApproach Type=Proposed2025.10 | 40 | 31.5 | 42.5 | 54.8 | 27.4 | 19.5 | 29.7 | 32.1 | 33 | |
| GeoDiffusionApproach Type=Bias Agnostic2025.10 | 38.3 | 28.4 | 39.6 | 52.4 | 25.3 | 18.3 | 27.6 | 30.5 | 32.1 | |
| GeoDiffusion + ResamplingApproach Type=Frequency Aware, Resampling Strategy=Gupta et al., 20192025.10 | 38.3 | 28.8 | 40 | 52 | 25.4 | 18 | 27.5 | 30.8 | 32.3 | |
| Copy PasteApproach Type=Bias Agnostic2025.10 | 37.5 | 28.6 | 38.8 | 51.5 | 25.3 | 16 | 24.7 | 31.5 | 32.7 | |
| Faster R-CNNConfiguration=Baseline2025.10 | 36.9 | 27.9 | 38.5 | 50.7 | 25.1 | 15.5 | 24 | 31.3 | 32.5 | |
| ControlNet + ResamplingApproach Type=Frequency Aware, Resampling Strategy=Gupta et al., 20192025.10 | 36.5 | 27.9 | 38.5 | 51 | 24.5 | 13.6 | 24.2 | 30.4 | 31.9 | |
| ControlNetApproach Type=Bias Agnostic2025.10 | 36.4 | 27.6 | 38.3 | 51.2 | 24.4 | 13.6 | 24.1 | 30.3 | 31.8 |