3D Object Detection on DAIR-V2X
84.5AP@0.50SparseAlign
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SparseAlignBW(Mb)=< 1.32025.03 | 84.5 | — | 68.5 | — | |
| CoSDHFusion Strategy=intermediate-late fusion2025.10 | 79.99 | — | 66.01 | 84.59 | |
| FadeLeadFusion Strategy=intermediate-late fusion2025.10 | 79.66 | — | 66.51 | 83.98 | |
| WhisperNetParams=7.28 M2026.03 | 79.15 | — | 64.8 | — | |
| WaveCommSensor Modality=LiDAR-based2026.03 | 79 | 20 | — | 83.1 | |
| DI-V2XBW(Mb)=48.82025.03 | 78.8 | — | 66.2 | — | |
| DSRCParams=10.14 M2026.03 | 78.52 | — | 63.6 | — | |
| Where2CommFusion Strategy=intermediate-late fusion2025.10 | 78.29 | — | 64.03 | 83.26 | |
| CoAlignParams=11.42 M2026.03 | 77.72 | — | 62.84 | — | |
| ERMVPParams=11.87 M2026.03 | 76.75 | — | 63.5 | — | |
| HM-ViTSensor Modality=LiDAR-based2026.03 | 76.1 | 23 | — | 81.8 | |
| MDDParams=8.09 M2026.03 | 74.95 | — | 58.17 | — | |
| CoAlignBW(Mb)=48.82025.03 | 74.6 | — | 60.4 | — | |
| CoBEVFlowNoise Level σt/σr(m/°)=0.1/0.1, Backbone=PointPillars2023.09 | 74.3 | — | 59.3 | — | |
| CoBEVFlowNoise Level σt/σr(m/°)=0.0/0.0, Backbone=PointPillars2023.09 | 73.8 | — | 59.9 | — | |
| DiscoNetBW(Mb)=48.82025.03 | 73.6 | — | 58.3 | — | |
| FcooperBW(Mb)=48.82025.03 | 73.4 | — | 55.9 | — | |
| OPV2VBW(Mb)=97.62025.03 | 73.3 | — | 55.3 | — | |
| CoBEVFlowNoise Level σt/σr(m/°)=0.2/0.2, Backbone=PointPillars2023.09 | 73.2 | — | 57.9 | — | |
| CoRAPose error Level (σt/σr)=0.6/0.6, Comm. (MB)=2.842025.12 | 73.12 | — | 58.03 | — | |
| EIMCMix-Voxel module=true2026.03 | 73.01 | 20.16 | 58.37 | 75.01 | |
| SparseAlign2025.03 | 72.7 | — | 35.2 | — | |
| BM2CP2026.03 | 72.37 | 23.18 | 56.18 | 77.91 | |
| CoBEVFlowNoise Level σt/σr(m/°)=0.3/0.3, Backbone=PointPillars2023.09 | 72.3 | — | 57.1 | — | |
| CATNetPublished=–, Params=09.95 M2026.03 | 72.3 | — | 56.5 | — | |
| V2VNet2026.03 | 72.22 | 25.43 | 52.95 | 77.36 | |
| V2X-VIT2026.03 | 71.87 | 24 | 55.46 | 77.29 | |
| EIMCMix-Voxel module=false2026.03 | 71.81 | 18.97 | 56.76 | 76.82 | |
| CoBEVT2026.03 | 71.7 | 24 | 55.85 | 77.86 | |
| Where2comm+SyncNetNoise Level σt/σr(m/°)=0.0/0.0, Backbone=PointPillars2023.09 | 71.1 | — | 58.7 | — | |
| CoAlignPose error Level (σt/σr)=0.6/0.6, Comm. (MB)=11.422025.12 | 70.64 | — | 57.52 | — | |
| V2X-ViTParams=14.50 M2026.03 | 70.46 | — | 52.4 | — | |
| CoSDHParams=8.52 M2026.03 | 70.42 | — | 57.66 | — | |
| V2XViTBW(Mb)=48.82025.03 | 70.4 | — | 53.1 | — | |
| CoBEVFlowNoise Level σt/σr(m/°)=0.4/0.4, Backbone=PointPillars2023.09 | 70.3 | — | 56 | — | |
| Where2commNoise Level σt/σr(m/°)=0.0/0.0, Backbone=PointPillars2023.09 | 70.2 | — | 57.7 | — | |
| DSRCPublished=AAAI’25, Params=40.64 M, Category=Noise-robust2026.03 | 70.2 | — | 55.9 | — | |
| LateParams=6.28 M2026.03 | 69.71 | — | 51.07 | — | |
| V2X-ViTPublished=ICRA’22, Params=13.50 M, Category=Noise-robust2026.03 | 69.6 | — | 51.7 | — | |
| HEAL (Pyramid Fusion)Modality=L2026.03 | 69.35 | — | 42.74 | 77.06 | |
| V2X-ViTNoise Level σt/σr(m/°)=0.0/0.0, Backbone=PointPillars2023.09 | 69.3 | — | 54.5 | — | |
| Where2commNoise Level σt/σr(m/°)=0.1/0.1, Backbone=PointPillars2023.09 | 69.3 | — | 57.7 | — | |
| CoAlignPublished=ICRA’23, Params=11.43 M, Category=Noise-robust2026.03 | 69.3 | — | 54.9 | — | |
| Fcooper2026.03 | 69.29 | 24 | 51.37 | 76.61 | |
| V2X-ViTNoise Level σt/σr(m/°)=0.1/0.1, Backbone=PointPillars2023.09 | 69.2 | — | 54.5 | — | |
| Where2comm+SyncNetNoise Level σt/σr(m/°)=0.1/0.1, Backbone=PointPillars2023.09 | 69.2 | — | 58.3 | — | |
| CoBEVTSensor Modality=LiDAR-based2026.03 | 69.2 | 23 | — | 78.7 | |
| V2X-ViTNoise Level σt/σr(m/°)=0.3/0.3, Backbone=PointPillars2023.09 | 68.5 | — | 68.5 | — | |
| DiscoNetSensor Modality=LiDAR-based2026.03 | 68.5 | 23 | — | 74.6 | |
| V2X-ViTNoise Level σt/σr(m/°)=0.4/0.4, Backbone=PointPillars2023.09 | 68.1 | — | 54.3 | — | |
| Where2commNoise Level σt/σr(m/°)=0.2/0.2, Backbone=PointPillars2023.09 | 67.9 | — | 56.1 | — | |
| DSRCPose error Level (σt/σr)=0.6/0.6, Comm. (MB)=11.422025.12 | 67.54 | — | 54.25 | — | |
| ERMVPPublished=CVPR’24, Params=12.42 M, Category=Latency-aware2026.03 | 67.4 | — | 55.4 | — | |
| Where2commParams=10.43 M2026.03 | 67.35 | — | 53.17 | — | |
| AttFusionSensor Modality=LiDAR-based2026.03 | 67.3 | 23 | — | 73.8 | |
| Where2comm+SyncNetNoise Level σt/σr(m/°)=0.4/0.4, Backbone=PointPillars2023.09 | 67.1 | — | 56.7 | — | |
| FPVRCNNBW(Mb)=0.242025.03 | 66.5 | — | 50.5 | — | |
| MRCNetPublished=CVPR’24, Params=19.71 M, Category=Latency-aware2026.03 | 66.5 | — | 53.9 | — | |
| MRCNetParams=18.79 M2026.03 | 66.47 | — | 53.85 | — | |
| V2VNetParams=16.08 M2026.03 | 66.44 | — | 40.37 | — | |
| V2VNetBW(Mb)=48.82025.03 | 66.4 | — | 40.2 | — | |
| CoSDHPose error Level (σt/σr)=0.6/0.6, Comm. (MB)=3.222025.12 | 66.07 | — | 52.67 | — | |
| V2X-ViTPose error Level (σt/σr)=0.6/0.6, Comm. (MB)=11.422025.12 | 66.01 | — | 50.36 | — | |
| Where2commNoise Level σt/σr(m/°)=0.3/0.3, Backbone=PointPillars2023.09 | 65.8 | — | 65.8 | — | |
| Where2commPublished=NIPS’22, Params=11.43 M, Category=Regular2026.03 | 65.4 | — | 50.2 | — | |
| Inter.Params=6.28 M2026.03 | 65.06 | — | 55.38 | — | |
| No Coll2026.03 | 65.02 | 0 | 53.82 | 69.99 | |
| SiMO (Pyramid Fusion)Modality=L+C (w/o RD)2026.03 | 64.51 | — | 29.31 | 74.56 | |
| MDDPose error Level (σt/σr)=0.6/0.6, Comm. (MB)=17.132025.12 | 64.37 | — | 49.82 | — | |
| Where2commNoise Level σt/σr(m/°)=0.4/0.4, Backbone=PointPillars2023.09 | 64.3 | — | 54.3 | — | |
| Where2commComm Volume (log2)=21.782022.09 | 63.76 | 21.78 | 48.94 | — | |
| Where2commComm Volume (log2)=22.352022.09 | 63.71 | 22.35 | 48.89 | — | |
| Where2commComm Volume (log2)=22.622022.09 | 63.71 | 22.62 | 48.93 | — | |
| Where2commComm Volume (log2)=20.562022.09 | 63.54 | 20.56 | 48.78 | — | |
| CoMambaPublished=Arxiv’24, Params=09.15 M, Category=Regular2026.03 | 63.5 | — | 47.9 | — | |
| Attn2026.03 | 63.16 | 24 | 49.17 | 68.78 | |
| Where2comm2026.03 | 63.16 | 24 | 51.04 | 68.18 | |
| SinglePose error Level (σt/σr)=0/02025.12 | 62.5 | — | 44.57 | — | |
| SingleParams=6.28 M2026.03 | 62.5 | — | 44.57 | — | |
| No FusionPublished=–, Params=–2026.03 | 62.5 | — | 44.6 | — | |
| How2commPublished=NIPS’23, Params=35.80 M, Category=Latency-aware2026.03 | 62.4 | — | 47.2 | — | |
| How2CommParams=35.79 M2026.03 | 62.36 | — | 47.18 | — | |
| ERMVPPose error Level (σt/σr)=0.6/0.6, Comm. (MB)=3.432025.12 | 62.28 | — | 52.14 | — | |
| F-CooperSensor Modality=LiDAR-based2026.03 | 62 | 23 | — | 72.3 | |
| V2VNetPose error Level (σt/σr)=0.6/0.6, Comm. (MB)=11.712025.12 | 59.84 | — | 35.08 | — | |
| TransIFF2025.03 | 59.6 | — | 46 | — | |
| MRCNetPose error Level (σt/σr)=0.6/0.6, Comm. (MB)=11.422025.12 | 59.35 | — | 47.77 | — | |
| Where2commComm Volume (log2)=18.192022.09 | 58.46 | 18.19 | 44.46 | — | |
| COREFusion Strategy=intermediate-late fusion2025.10 | 58.32 | — | 46.82 | 62.14 | |
| Where2comm+SyncNetNoise Level σt/σr(m/°)=0.2/0.2, Backbone=PointPillars2023.09 | 58.3 | — | 57.9 | — | |
| Where2comm+SyncNetNoise Level σt/σr(m/°)=0.3/0.3, Backbone=PointPillars2023.09 | 57.9 | — | 57 | — | |
| V2VNet2022.09 | 56.01 | 24.21 | 42.25 | — | |
| Where2commComm Volume (log2)=17.532022.09 | 55.84 | 17.53 | 42.44 | — | |
| V2X-ViTNoise Level σt/σr(m/°)=0.2/0.2, Backbone=PointPillars2023.09 | 54.5 | — | 54.5 | — | |
| DiscoNet2022.09 | 54.29 | 22.62 | 44.88 | — | |
| V2X-ViT2022.09 | 54.26 | 22.62 | 43.35 | — | |
| Where2commComm Volume (log2)=17.032022.09 | 53.53 | 17.03 | 40.7 | — | |
| Late Fusion2022.09 | 53.12 | 11.45 | 37.88 | — | |
| SiMO (Pyramid Fusion)Modality=L (RD)2026.03 | 52.33 | — | 19.35 | 66.89 | |
| V2X-ViTSensor Modality=LiDAR-based2026.03 | 52.1 | 23 | — | 78.5 |