3D Object Detection on aiMotive (test)
66.7Mean AP (all-point)HyperDUM
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| HyperDUMFusion Architecture=BEVFusion, UQ Method=HyperDUM2025.03 | 66.7 | 72.23 | 69.58 | 64.77 | 64.48 | 76.69 | 75.15 | 44.78 | 45.48 | 66 | |
| BEVFusionFusion Architecture=BEVFusion, UQ Method=None (Baseline)2025.03 | 65.67 | 72.55 | 69.08 | 63.72 | 63.52 | 74.74 | 73.13 | 42.75 | 42.83 | 64.79 | |
| HyperDUMUQ Method=HyperDUM2025.03 | 65.16 | — | — | — | — | — | — | — | — | 64.62 | |
| LDUFusion Architecture=BEVFusion, UQ Method=Latent Deterministic Uncertainty, number of learnable prototypes=42025.03 | 64.69 | 71.48 | 69.33 | 62.54 | 60.11 | 76.49 | 75.04 | 40.49 | 41.24 | 64.73 | |
| InfNoiseFusion Architecture=BEVFusion, Number of forward passes=10, UQ Method=infer-noise2025.03 | 64.59 | 70.11 | 68.52 | 63.1 | 60.42 | 75.08 | 74.19 | 35.58 | 37.41 | 64.56 | |
| InfMCDFusion Architecture=BEVFusion, Number of forward passes=10, UQ Method=Infer-dropout2025.03 | 64.56 | 72.58 | 69.65 | 62.02 | 59.59 | 75.05 | 73.72 | 35.58 | 37.41 | 64.58 | |
| InfNoiseUQ Method=InfNoise2025.03 | 63.98 | — | — | — | — | — | — | — | — | 63.1 | |
| LDUUQ Method=LDU2025.03 | 63.87 | — | — | — | — | — | — | — | — | 63.05 | |
| BEVFusionUQ Method=BEVFusion2025.03 | 63.79 | — | — | — | — | — | — | — | — | 62.75 | |
| InfMCDUQ Method=InfMCD2025.03 | 63.5 | — | — | — | — | — | — | — | — | 63.39 | |
| PostNetFusion Architecture=BEVFusion, UQ Method=Posterior Network2025.03 | 63.13 | 69.19 | 68.02 | 61.38 | 59.84 | 73.32 | 68.97 | 34.95 | 37.6 | 60.21 | |
| PostNetUQ Method=PostNet2025.03 | 61.97 | — | — | — | — | — | — | — | — | 59.69 |