3D Object Detection on KITTI (val) (IoU 0.5 & 0.7)
86.83AP3D (Moderate)SA-SSD+EBM
Evaluation results
| Method | Links | |||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SA-SSD+EBMrefinement=gradient-based refinement (Algorithm 1)2020.12 | 86.83 | — | — | — | — | — | — | 95.45 | 82.23 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLOCS-PVCas2020.12 | 85.94 | — | — | — | — | — | — | 92.78 | 83.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LSK3DNetInput=V2024.03 | 85.61 | — | — | — | — | — | — | 90.16 | 79.53 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LargeKernel3DInput=V2024.03 | 85.07 | — | — | — | — | — | — | 89.52 | 79.32 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Focals ConvInput=V2024.03 | 84.93 | — | — | — | — | — | — | 89.52 | 79.18 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PV-RCNN2020.12 | 84.83 | — | — | — | — | — | — | 92.57 | 82.69 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SA-SSDmodel_source=pre-trained model provided by authors2020.12 | 84.65 | — | — | — | — | — | — | 93.14 | 81.86 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Voxel R-CNNInput=V2024.03 | 84.52 | — | — | — | — | — | — | 89.41 | 78.93 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Voxel R-CNNRecall positions=11, Category=Car2021.09 | 84.52 | — | — | — | — | — | — | 89.41 | 78.93 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SA-SSDresults_source=reported in original paper [23]2020.12 | 84.3 | — | — | — | — | — | — | 93.23 | 81.36 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoTr-TSDRecall positions=11, Category=Car2021.09 | 84.04 | — | — | — | — | — | — | 89.04 | 78.68 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PV-RCNNInput=PV2024.03 | 83.69 | — | — | — | — | — | — | 89.35 | 78.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PV-RCNNRecall positions=11, Category=Car2021.09 | 83.69 | — | — | — | — | — | — | 89.35 | 78.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SASSD2020.12 | 79.91 | — | — | — | — | — | — | 90.15 | 78.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SA-SSDRecall positions=11, Category=Car2021.09 | 79.91 | — | — | — | — | — | — | 90.15 | 78.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CIA-SSD2020.12 | 79.81 | — | — | — | — | — | — | 90.04 | 78.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STDRecall positions=11, Category=Car2021.09 | 79.8 | — | — | — | — | — | — | 89.7 | 79.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Part-A²Input=V2024.03 | 79.47 | — | — | — | — | — | — | 89.47 | 78.54 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Part-A2 NetRecall positions=11, Category=Car2021.09 | 79.47 | — | — | — | — | — | — | 89.47 | 78.54 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 3DSSD2020.12 | 79.45 | — | — | — | — | — | — | 89.71 | 78.67 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 3DSSDRecall positions=11, Category=Car2021.09 | 79.45 | — | — | — | — | — | — | 89.71 | 78.67 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Associate-3Ddet2020.12 | 79.17 | — | — | — | — | — | — | 89.29 | 77.76 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Frustum ConvNet2019.03 | 78.8 | — | — | — | — | — | — | 89.02 | 77.09 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointRCNN2019.03 | 78.63 | — | — | — | — | — | — | 88.88 | 77.38 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point R-CNNInput=P2024.03 | 78.63 | — | — | — | — | — | — | 88.88 | 77.38 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointRCNNRecall positions=11, Category=Car2021.09 | 78.63 | — | — | — | — | — | — | 88.88 | 77.38 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SECONDInput=V2024.03 | 78.62 | — | — | — | — | — | — | 88.61 | 77.22 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-GNN2020.12 | 78.34 | — | — | — | — | — | — | 87.89 | 77.38 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoTr-SSDRecall positions=11, Category=Car2021.09 | 78.27 | — | — | — | — | — | — | 87.86 | 76.93 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointPillars2020.12 | 77.98 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TANet2020.12 | 77.85 | — | — | — | — | — | — | 88.21 | 75.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TANetRecall positions=11, Category=Car2021.09 | 76.64 | — | — | — | — | — | — | 87.52 | 73.86 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SECOND2020.12 | 76.48 | — | — | — | — | — | — | 87.43 | 69.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SECONDRecall positions=11, Category=Car2021.09 | 76.48 | — | — | — | — | — | — | 87.43 | 69.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| IPOD2019.03 | 76.4 | — | — | — | — | — | — | 84.1 | 75.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointPillarsInput=R2024.03 | 76.06 | — | — | — | — | — | — | 86.62 | 68.91 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointPillarsRecall positions=11, Category=Car2021.09 | 76.06 | — | — | — | — | — | — | 86.62 | 68.91 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AVOD-FPN2019.03 | 74.44 | — | — | — | — | — | — | 84.41 | 68.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ContFusion2019.03 | 73.25 | — | — | — | — | — | — | 86.32 | 67.81 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ContFuse2020.12 | 73.25 | — | — | — | — | — | — | 86.32 | 67.81 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Ours (PV-RCNN)Detector=PV-RCNN, Setting=AD-PT, Data amount=100%2023.06 | 73.01 | — | — | — | — | — | — | — | — | — | — | — | 68.87 | 60.79 | 55.42 | 91.81 | 73.49 | 69.21 | — | 91.96 | 84.75 | 82.53 | — | — | — | |
| Proposal Contrast (PV-RCNN)Detector=PV-RCNN, Setting=SS-PT, Data amount=100%2023.06 | 72.92 | — | — | — | — | — | — | — | — | — | — | — | 68.43 | 60.36 | 55.01 | 92.77 | 73.69 | 69.51 | — | 92.45 | 84.72 | 82.47 | — | — | — | |
| PointContrast (PV-RCNN)Detector=PV-RCNN, Setting=SS-PT, Data amount=100%2023.06 | 71.55 | — | — | — | — | — | — | — | — | — | — | — | 65.73 | 57.74 | 52.46 | 91.47 | 72.72 | 67.95 | — | 91.4 | 84.18 | 82.25 | — | — | — | |
| STRL (PV-RCNN)Detector=PV-RCNN, Setting=SS-PT, Data amount=100%2023.06 | 71.46 | — | — | — | — | — | — | — | — | — | — | — | — | 57.8 | — | — | 71.88 | — | — | — | 84.7 | — | — | — | — | |
| GCC-3D (PV-RCNN)Detector=PV-RCNN, Setting=SS-PT, Data amount=100%2023.06 | 71.26 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| F-PointNet2019.03 | 70.92 | — | — | — | — | — | — | 83.76 | 63.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| From scratch (PV-RCNN)Detector=PV-RCNN, Setting=From scratch, Data amount=100%2023.06 | 70.57 | — | — | — | — | — | — | — | — | — | — | — | — | 57.06 | — | — | 70.14 | — | — | — | 84.5 | — | — | — | — | |
| Ours (PV-RCNN)Detector=PV-RCNN, Setting=AD-PT, Data amount=20%2023.06 | 69.43 | — | — | — | — | — | — | — | — | — | — | — | 65.5 | 57.59 | 51.84 | 84.15 | 67.96 | 64.73 | — | 92.18 | 82.75 | 82.12 | — | — | — | |
| Proposal Contrast (PV-RCNN)Detector=PV-RCNN, Setting=SS-PT, Data amount=20%2023.06 | 68.13 | — | — | — | — | — | — | — | — | — | — | — | 62.58 | 55.05 | 50.06 | 88.58 | 66.68 | 62.32 | — | 91.96 | 82.65 | 80.15 | — | — | — | |
| Ours (SECOND)Detector=SECOND, Setting=AD-PT, Data amount=100%2023.06 | 67.58 | — | — | — | — | — | — | — | — | — | — | — | 58.3 | 53.58 | 48.72 | 86.04 | 67.78 | 63.95 | — | 90.36 | 81.39 | 78.41 | — | — | — | |
| From scratch (PV-RCNN)Detector=PV-RCNN, Setting=From scratch, Data amount=20%2023.06 | 66.71 | — | — | — | — | — | — | — | — | — | — | — | 58.78 | 53.33 | 47.61 | 86.74 | 64.28 | 59.53 | — | 91.81 | 82.52 | 80.11 | — | — | — | |
| From scratch (SECOND)Detector=SECOND, Setting=From scratch, Data amount=100%2023.06 | 66.7 | — | — | — | — | — | — | — | — | — | — | — | 58.05 | 52.61 | 48.24 | 84.25 | 66.71 | 62.5 | — | 89.63 | 80.78 | 78.21 | — | — | — | |
| Ours (SECOND)Detector=SECOND, Setting=AD-PT, Data amount=20%2023.06 | 65.95 | — | — | — | — | — | — | — | — | — | — | — | 55.63 | 49.67 | 45.12 | 83.78 | 67.5 | 63.4 | — | 90.23 | 80.7 | 78.29 | — | — | — | |
| VoxelNet2019.03 | 65.46 | — | — | — | — | — | — | 81.97 | 62.85 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoxelNet2020.12 | 65.46 | — | — | — | — | — | — | 81.97 | 62.85 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoxelNetInput=V2024.03 | 65.46 | — | — | — | — | — | — | 81.97 | 62.85 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoxelNetRecall positions=11, Category=Car2021.09 | 65.46 | — | — | — | — | — | — | 81.97 | 62.85 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STONEBackbone detection model=SECOND [61], Query budget=1%2024.10 | 64.04 | — | — | — | — | — | — | 76.86 | 58.75 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MV-JARInitialization=MV-JAR, Backbone=SST, Pre-training dataset=Waymo, Fine-tuning dataset=KITTI2023.03 | 63.8 | — | — | — | — | — | — | 75.22 | 60.35 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandomInitialization=Random, Backbone=SST, Pre-training dataset=None, Fine-tuning dataset=KITTI2023.03 | 63.43 | — | — | — | — | — | — | 74.71 | 60 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ProposalContrastInitialization=ProposalContrast, Backbone=SST, Pre-training dataset=Waymo, Fine-tuning dataset=KITTI2023.03 | 63.34 | — | — | — | — | — | — | 73.63 | 59.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| High-ResolutionDescription=Ground-truth data, Downstream Model=PointPillars2023.12 | 62.78 | — | — | — | — | — | — | 73.33 | 59.63 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MV3D2019.03 | 62.68 | — | — | — | — | — | — | 71.29 | 56.56 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointContrastInitialization=PointContrast, Backbone=SST, Pre-training dataset=Waymo, Fine-tuning dataset=KITTI2023.03 | 62.53 | — | — | — | — | — | — | 73.35 | 59.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| From scratch (SECOND)Detector=SECOND, Setting=From scratch, Data amount=20%2023.06 | 61.7 | — | — | — | — | — | — | — | — | — | — | — | 52.08 | 47.23 | 43.37 | 76.35 | 59.06 | 55.24 | — | 89.78 | 78.83 | 76.21 | — | — | — | |
| KECORBackbone detection model=SECOND [61], Query budget=1%2024.10 | 60.38 | — | — | — | — | — | — | 74.05 | 55.34 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CRBBackbone detection model=SECOND [61], Query budget=1%2024.10 | 58.06 | — | — | — | — | — | — | 72.33 | 53.09 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BAITBackbone detection model=SECOND [61], Query budget=1%2024.10 | 55.61 | — | — | — | — | — | — | 69.45 | 51.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BADGEBackbone detection model=SECOND [61], Query budget=1%2024.10 | 55.6 | — | — | — | — | — | — | 69.92 | 51.23 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RandomBackbone detection model=SECOND [61], Query budget=1%2024.10 | 55.48 | — | — | — | — | — | — | 66.33 | 51.53 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLALBackbone detection model=SECOND [61], Query budget=1%2024.10 | 55.38 | — | — | — | — | — | — | 69.19 | 50.85 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CORESETBackbone detection model=SECOND [61], Query budget=1%2024.10 | 53.22 | — | — | — | — | — | — | 66.86 | 48.97 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TULIP-LUpsampling Ratio=x4, Downstream Model=PointPillars2023.12 | 41.33 | — | — | — | — | — | — | 54.15 | 37.19 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TULIPUpsampling Ratio=x4, Downstream Model=PointPillars2023.12 | 37.57 | — | — | — | — | — | — | 50.23 | 32.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ILNUpsampling Ratio=x4, Downstream Model=PointPillars2023.12 | 28.61 | — | — | — | — | — | — | 38.29 | 23.67 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Pseudo-Stereovariant=Ours-fld (feature-level generation)2022.03 | 24.15 | — | — | — | — | — | — | 35.18 | 20.35 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LiDAR-SRUpsampling Ratio=x4, Downstream Model=PointPillars2023.12 | 24.15 | — | — | — | — | — | — | 29.27 | 20.39 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Pseudo-Stereovariant=Ours-im (image-level generation)2022.03 | 22.36 | — | — | — | — | — | — | 31.81 | 19.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDMP-3D2022.03 | 20.39 | — | — | — | — | — | — | 28.12 | 16.34 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Pseudo-Stereovariant=Ours-fcd (feature-clone)2022.03 | 19.15 | — | — | — | — | — | — | 28.46 | 16.56 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MonoFlex2022.03 | 17.51 | — | — | — | — | — | — | 23.64 | 14.83 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GUPNet2022.03 | 16.46 | — | — | — | — | — | — | 22.76 | 13.72 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CaDDN2022.03 | 16.31 | — | — | — | — | — | — | 23.57 | 13.84 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| D4LCN2022.03 | 16.2 | — | — | — | — | — | — | 22.32 | 12.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Low-ResolutionDescription=Input low-resolution data, Downstream Model=PointPillars2023.12 | 9.03 | — | — | — | — | — | — | 10.05 | 8.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 3DOPSensor=Stereo2019.02 | — | 46.04 | 34.63 | 30.09 | 6.55 | 5.07 | 4.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 3DOPInput=Stereo2019.04 | — | 46.04 | 34.63 | 30.09 | 6.55 | 5.07 | 4.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 3DOPExtra data=Stereo, Inference Time=3 s2020.01 | — | 46.04 | 34.63 | 30.09 | 6.55 | 5.07 | 4.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 3DSSDModality=LiDAR2023.01 | — | — | — | — | — | — | — | — | — | 88.55 | 78.45 | 77.3 | 58.18 | 54.31 | 49.56 | 86.25 | 70.48 | 65.32 | 69.82 | — | — | — | — | — | — | |
| AVOD-FPNModality=LiDAR+RGB2023.01 | — | — | — | — | — | — | — | — | — | 84.41 | 74.44 | 68.65 | — | 58.8 | — | — | 49.7 | — | — | — | — | — | — | — | — | |
| CAT-DetModality=LiDAR+RGB2023.01 | — | — | — | — | — | — | — | — | — | 90.12 | 81.46 | 79.15 | 74.08 | 66.35 | 58.92 | 87.64 | 72.82 | 68.2 | 75.42 | — | — | — | — | — | — | |
| CenterNet2021.07 | — | 20 | 17.5 | 15.57 | 60 | 66 | 77 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CenterpointModality=L, Setting=Fully Sup., Label ratio=100%2025.12 | — | — | — | — | — | — | — | — | — | 89.07 | 80.5 | 76.49 | — | — | — | — | — | — | — | — | — | — | 94.77 | 52.11 | 69.32 | |
| CenterPointLabels=100%, Training Dataset=WOD2024.04 | — | 97.07 | 89.23 | 81.81 | 88.55 | 78.38 | 71.43 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLOCSModality=LiDAR+RGB2023.01 | — | — | — | — | — | — | — | — | — | 89.49 | 79.31 | 77.36 | 62.88 | 56.2 | 50.1 | 87.57 | 67.92 | 63.67 | 70.5 | — | — | — | — | — | — | |
| CoInModality=L, Setting=Weakly Sup., Label ratio=2%2025.12 | — | — | — | — | — | — | — | — | — | 89.17 | 75.32 | 62.98 | — | — | — | — | — | — | — | — | — | — | 83.89 | 31.69 | 49.1 | |
| CPDLabels=0, Training Dataset=WOD2024.04 | — | 90.85 | 81.01 | 79.8 | 72.98 | 55.07 | 53.94 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CPDModality=L, Setting=Unsup., Label ratio=0%2025.12 | — | — | — | — | — | — | — | — | — | 72.98 | 55.07 | 53.94 | — | — | — | — | — | — | — | — | — | — | 83.89 | 15.48 | 8.3 | |
| Deep3DBoxSensor=Mono2019.02 | — | 27.04 | 20.55 | 15.88 | 5.85 | 4.1 | 3.84 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Deep3DBoxInput=Monocular2019.04 | — | 27.04 | 20.55 | 15.88 | 5.85 | 4.1 | 3.84 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |