3D Object Detection on V2X-Real Gaussian Pose Noise σ = 0.4 (test)
84.27AP@0.3 (Vehicle)HyDRA w AG-PGO
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HyDRA w AG-PGOGaussian Pose Noise=σ = 0.4, Architecture Heterogeneity Setting=EP + HP + XS + XV2026.03 | 84.27 | 16.62 | 51.89 | 50.93 | 71.16 | 3.61 | 42.35 | 39.04 | 29.17 | 0.07 | 35.69 | 21.64 | |
| HyDRA w AG-PGOArchitecture=EP + HP + XS + XV, Pose Noise=Gaussian (sigma=0.4), Pose Correction=AG-PGO (Ours)2026.03 | 84.27 | 16.62 | 51.89 | 50.93 | 71.16 | 3.61 | 42.35 | 39.04 | 29.17 | 0.07 | 35.69 | 21.64 | |
| E2E TrainingGaussian Pose Noise=σ = 0.4, Architecture Heterogeneity Setting=EP + HP + XS + XV2026.03 | 83.9 | 9.51 | 52.13 | 48.51 | 72.57 | 1.53 | 45.97 | 40.02 | 25.8 | 0.01 | 22.5 | 16.1 | |
| E2E TrainingArchitecture=EP + HP + XS + XV, Pose Noise=Gaussian (sigma=0.4)2026.03 | 83.9 | 9.51 | 52.13 | 48.51 | 72.57 | 1.53 | 45.97 | 40.02 | 25.8 | 0.01 | 22.5 | 16.1 | |
| HEALGaussian Pose Noise=σ = 0.4, Architecture Heterogeneity Setting=EP + HP + XS + XV2026.03 | 83.76 | 11.65 | 53.27 | 49.65 | 73.12 | 1.94 | 37.01 | 37.36 | 31.07 | 0.01 | 17.51 | 16.2 | |
| HEALArchitecture=EP + HP + XS + XV, Pose Noise=Gaussian (sigma=0.4)2026.03 | 83.76 | 11.65 | 53.27 | 49.65 | 73.12 | 1.94 | 37.01 | 37.36 | 31.07 | 0.01 | 17.51 | 16.2 | |
| HyDRA w CoAlignGaussian Pose Noise=σ = 0.4, Architecture Heterogeneity Setting=EP + HP + XS + XV2026.03 | 82.06 | 12.63 | 50.16 | 48.28 | 66.57 | 2.44 | 36.31 | 35.11 | 18.24 | 0.04 | 21.81 | 13.36 | |
| HyDRA w CoAlignArchitecture=EP + HP + XS + XV, Pose Noise=Gaussian (sigma=0.4), Pose Correction=CoAlign [19]2026.03 | 82.06 | 12.63 | 50.16 | 48.28 | 66.57 | 2.44 | 36.31 | 35.11 | 18.24 | 0.04 | 21.81 | 13.36 | |
| MPDAGaussian Pose Noise=σ = 0.4, Architecture Heterogeneity Setting=EP + HP + XS + XV2026.03 | 80.96 | 15.36 | 54.12 | 50.14 | 75.33 | 3.06 | 51.46 | 43.28 | 41.15 | 0.07 | 34.12 | 25.11 | |
| MPDAArchitecture=EP + HP + XS + XV, Pose Noise=Gaussian (sigma=0.4)2026.03 | 80.96 | 15.36 | 54.12 | 50.14 | 75.33 | 3.06 | 51.46 | 43.28 | 41.15 | 0.07 | 34.12 | 25.11 | |
| GenCommGaussian Pose Noise=σ = 0.4, Architecture Heterogeneity Setting=EP + HP + XS + XV2026.03 | 80.6 | 11.04 | 50.9 | 47.51 | 74.71 | 2.13 | 46.81 | 41.22 | 34.75 | 0.02 | 28.13 | 20.97 | |
| GenCommArchitecture=EP + HP + XS + XV, Pose Noise=Gaussian (sigma=0.4)2026.03 | 80.6 | 11.04 | 50.9 | 47.51 | 74.71 | 2.13 | 46.81 | 41.22 | 34.75 | 0.02 | 28.13 | 20.97 | |
| CodeFillingGaussian Pose Noise=σ = 0.4, Architecture Heterogeneity Setting=EP + HP + XS + XV2026.03 | 75.9 | 10.88 | 49.41 | 45.4 | 64.58 | 1.96 | 42.84 | 36.46 | 21.42 | 0.02 | 15.61 | 12.35 | |
| CodeFillingArchitecture=EP + HP + XS + XV, Pose Noise=Gaussian (sigma=0.4)2026.03 | 75.9 | 10.88 | 49.41 | 45.4 | 64.58 | 1.96 | 42.84 | 36.46 | 21.42 | 0.02 | 15.61 | 12.35 | |
| Late FusionGaussian Pose Noise=σ = 0.4, Architecture Heterogeneity Setting=EP + HP + XS + XV2026.03 | 52.44 | 23.9 | 37.19 | 37.84 | 35.68 | 10.84 | 19.11 | 21.88 | 6.91 | 0.46 | 6.85 | 4.74 | |
| Late FusionArchitecture=EP + HP + XS + XV, Pose Noise=Gaussian (sigma=0.4)2026.03 | 52.44 | 23.9 | 37.19 | 37.84 | 35.68 | 10.84 | 19.11 | 21.88 | 6.91 | 0.46 | 6.85 | 4.74 |