3D Object Detection on KITTI Car (test)
1,257AP3D (Easy)MonoPSR
Evaluation results
| Method | Links | ||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MonoPSRRuntime=0.20s2019.04 | 1,257 | 1,085 | 906 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MultiFusionRuntime=0.12s2019.04 | 708 | 518 | 468 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| A3DODWTDARuntime=0.80s2019.04 | 676 | 645 | 487 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointViT V1Method Architecture=One-stage, Latency (ms)=36, Evaluation Hardware=1x RTX 4090 GPU, Batch Size=1, Depth Completion Fusion Strategy=Early Fusion + PE-Net2026.02 | 91.16 | 79.93 | 72.51 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Voxel-RCNN2023.03 | 90.9 | 81.62 | 77.06 | — | — | — | — | — | — | — | — | — | — | — | — | — | 83.19 | — | — | — | |
| OcTr2023.03 | 90.88 | 82.64 | 77.77 | — | — | — | — | — | — | — | — | — | — | — | — | — | 83.76 | — | — | — | |
| PVT-SSDMethod Architecture=One-stage, Latency (ms)=32, Evaluation Hardware=1x RTX 4090 GPU, Batch Size=12026.02 | 90.65 | 82.29 | 76.85 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Focals Conv2023.03 | 90.55 | 82.28 | 77.59 | — | — | — | — | — | — | — | — | — | — | — | — | — | 83.47 | — | — | — | |
| SPGMethod Architecture=Two-stage, Latency (ms)=1562026.02 | 90.5 | 82.13 | 78.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PDVMethod Architecture=Two-stage, Latency (ms)=1352026.02 | 90.43 | 81.86 | 77.36 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EQ-PVRCNNMethod Architecture=Two-stage, Latency (ms)=2002026.02 | 90.31 | 82 | 77.53 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| M3DETRMethod Architecture=Two-stage2026.02 | 90.28 | 81.73 | 76.96 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PV-RCNN2023.03 | 90.25 | 81.43 | 76.82 | — | — | — | — | — | — | — | — | — | — | — | — | — | 82.83 | — | — | — | |
| PV-RCNNMethod Architecture=Two-stage, Latency (ms)=982026.02 | 90.25 | 81.43 | 76.82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PV-RCNN++Method Architecture=Two-stage, Latency (ms)=602026.02 | 90.14 | 81.88 | 77.15 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PVGNetMethod Architecture=One-stage2026.02 | 89.94 | 81.81 | 77.09 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoTR-TSD2023.03 | 89.9 | 82.09 | 79.14 | — | — | — | — | — | — | — | — | — | — | — | — | — | 83.71 | — | — | — | |
| VoTr-TSDMethod Architecture=Two-stage, Latency (ms)=2162026.02 | 89.9 | 82.09 | 79.14 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointViT V2Method Architecture=One-stage, Latency (ms)=36, Evaluation Hardware=1x RTX 4090 GPU, Batch Size=1, Depth Completion Fusion Strategy=Early Fusion + BP-Net2026.02 | 89.81 | 80.54 | 74.96 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointViT P1Method Architecture=One-stage, Latency (ms)=50, Evaluation Hardware=1x RTX 4090 GPU, Batch Size=1, Depth Completion Fusion Strategy=Early Fusion + BP-Net, LiDAR Input Type=colored raw LiDAR points2026.02 | 89.22 | 79.97 | 72.64 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PC-RGNNModality=LiDAR only2020.12 | 89.13 | 79.9 | 75.54 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| IA-SSDMethod Architecture=One-stage, Latency (ms)=442026.02 | 88.87 | 80.32 | 75.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SASAMethod Architecture=One-stage, Latency (ms)=632026.02 | 88.76 | 82.16 | 77.16 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SA-SSD2023.03 | 88.75 | 79.79 | 74.16 | — | — | — | — | — | — | — | — | — | — | — | — | — | 80.9 | — | — | — | |
| SA-SSDMethod Architecture=One-stage, Latency (ms)=402026.02 | 88.75 | 79.79 | 74.16 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoxSet2023.03 | 88.53 | 82.06 | 77.46 | — | — | — | — | — | — | — | — | — | — | — | — | — | 82.68 | — | — | — | |
| MMFModality=RGB+LiDAR2020.12 | 88.4 | 77.43 | 70.22 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Pyramid-PVMethod Architecture=Two-stage, Latency (ms)=1272026.02 | 88.39 | 82.08 | 77.49 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 3DSSDModality=LiDAR only2020.12 | 88.36 | 79.57 | 74.55 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 3DSSD2023.03 | 88.36 | 79.57 | 74.55 | — | — | — | — | — | — | — | — | — | — | — | — | — | 80.83 | — | — | — | |
| 3DSSDMethod Architecture=One-stage, Latency (ms)=382026.02 | 88.36 | 79.57 | 74.55 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-GNNModality=LiDAR only2020.12 | 88.33 | 79.47 | 72.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Point-GNNMethod Architecture=One-stage, Latency (ms)=6432026.02 | 88.33 | 79.47 | 72.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STDModality=LiDAR only2020.12 | 87.95 | 79.71 | 75.09 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STD2023.03 | 87.95 | 79.71 | 75.09 | — | — | — | — | — | — | — | — | — | — | — | — | — | 80.92 | — | — | — | |
| STDMethod Architecture=Two-stage, Latency (ms)=802026.02 | 87.95 | 79.71 | 75.09 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CT3D2023.03 | 87.83 | 81.77 | 77.16 | — | — | — | — | — | — | — | — | — | — | — | — | — | 82.25 | — | — | — | |
| CT3DMethod Architecture=Two-stage, Latency (ms)=702026.02 | 87.83 | 81.77 | 77.16 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Part-A2Modality=LiDAR only2020.12 | 87.81 | 78.49 | 73.51 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SVGA-NetMethod Architecture=One-stage, Latency (ms)=622026.02 | 87.33 | 80.47 | 75.91 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointRCNNModality=LiDAR only2020.12 | 86.96 | 75.64 | 70.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointRCNNMethod Architecture=Two-stage, Latency (ms)=1002026.02 | 86.96 | 75.64 | 70.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UberATG-MMFModality=RGB + LiDAR2019.07 | 86.81 | 76.75 | 68.41 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STDModality=LiDAR2019.07 | 86.61 | 77.63 | 76.06 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointRCNNModality=LiDAR2019.07 | 85.94 | 75.76 | 68.32 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Fast Point R-CNNModality=LiDAR only2020.12 | 85.29 | 77.4 | 70.24 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TANetMethod Architecture=One-stage, Latency (ms)=352026.02 | 84.39 | 75.94 | 68.82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RoarNetModality=RGB + LiDAR2019.07 | 83.95 | 75.79 | 67.88 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ContFuseModality=RGB+LiDAR2020.12 | 83.68 | 68.78 | 61.67 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SECOND2023.03 | 83.34 | 72.55 | 65.82 | — | — | — | — | — | — | — | — | — | — | — | — | — | 73.9 | — | — | — | |
| SECONDMethod Architecture=One-stage, Latency (ms)=502026.02 | 83.34 | 72.55 | 65.82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SECONDModality=LiDAR2019.07 | 83.13 | 73.66 | 66.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SECONDModality=LiDAR only2020.12 | 83.13 | 73.66 | 66.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AVOD-FPNModality=RGB+LiDAR2020.12 | 83.07 | 71.76 | 65.73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointPillarsModality=LiDAR only2020.12 | 82.58 | 74.31 | 68.99 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointPillars2023.03 | 82.58 | 74.31 | 68.99 | — | — | — | — | — | — | — | — | — | — | — | — | — | 75.29 | — | — | — | |
| PointPillarsMethod Architecture=One-stage, Latency (ms)=242026.02 | 82.58 | 74.31 | 68.99 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| F-PointNetModality=RGB+LiDAR2020.12 | 82.19 | 69.79 | 60.59 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AVOD (Feature Pyramid)Runtime (s)=0.12017.12 | 81.94 | 71.88 | 66.38 | 88.53 | 83.79 | 77.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AVOD-FPNModality=RGB + LiDAR2019.07 | 81.94 | 71.88 | 66.38 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AVODInput signal=LiDAR + Mono, IoU threshold=0.72018.12 | 81.9 | 71.9 | 66.4 | 88.5 | 83.8 | 77.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| F-PointNetRuntime (s)=0.172017.12 | 81.2 | 70.39 | 62.19 | 88.7 | 84 | 75.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| F-PointNetModality=RGB + LiDAR2019.07 | 81.2 | 70.39 | 62.19 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| F-POINTNETInput signal=LiDAR + Mono, IoU threshold=0.72018.12 | 81.2 | 70.4 | 62.2 | 88.7 | 84 | 75.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PointPillarsModality=LiDAR2019.07 | 79.05 | 74.99 | 68.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoxelNetRuntime (s)=0.232017.12 | 77.47 | 65.11 | 57.73 | 89.35 | 79.26 | 77.39 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoxelNetModality=LiDAR2019.07 | 77.47 | 65.11 | 57.73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoxelNetModality=LiDAR only2020.12 | 77.47 | 65.11 | 57.73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MV3DModality=RGB+LiDAR2020.12 | 74.97 | 63.63 | 54 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CDN-DSGNBase Method=DSGN, Feature=CDN-SDN2020.07 | 74.5 | 54.2 | 46.4 | 83.3 | 66.2 | 57.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AVODRuntime (s)=0.082017.12 | 73.59 | 65.78 | 58.38 | 86.8 | 85.44 | 77.73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AVODModality=RGB + LiDAR2019.07 | 73.59 | 65.78 | 58.38 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSGNTime=0.67s, Point cloud or pretrained disparity module=true2021.03 | 73.5 | 52.18 | 45.14 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSGN2020.07 | 73.5 | 52.2 | 45.1 | 82.9 | 65 | 56.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MV3DRuntime (s)=0.362017.12 | 71.09 | 62.35 | 55.12 | 86.02 | 76.9 | 68.49 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MV3DModality=RGB + LiDAR2019.07 | 71.09 | 62.35 | 55.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| YOLOStereo3DTime=0.08s, Point cloud or pretrained disparity module=false2021.03 | 65.68 | 41.25 | 30.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PSEUDO-LIDAR E2ETraining=End-to-end fine-tuning2020.07 | 64.8 | 43.9 | 38.1 | 79.6 | 58.8 | 52.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CDN-PSEUDO-LIDAR ++Base Method=PSEUDO-LIDAR ++, Feature=CDN-SDN2020.07 | 64.3 | 44.9 | 38.1 | 81.3 | 61 | 52.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Pseudo-LiDAR++Time=0.40s, Point cloud or pretrained disparity module=true2021.03 | 61.11 | 42.43 | 36.99 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PSEUDO-LIDAR ++2020.07 | 61.1 | 42.4 | 37 | 78.3 | 58 | 51.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DISP R-CNN2020.07 | 59.6 | 39.4 | 32 | 74.1 | 52.4 | 43.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Disp R-CNN(velo)Time=0.42s, Point cloud or pretrained disparity module=true2021.03 | 59.58 | 39.34 | 31.99 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ZoomNetTime=0.35s, Point cloud or pretrained disparity module=true2021.03 | 55.98 | 38.64 | 30.97 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AVODInput signal=Stereo, Representation=pseudo-LiDAR (PSMNet*), IoU threshold=0.72018.12 | 55.4 | 37.2 | 31.4 | 66.8 | 47.2 | 40.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OC-STEREO2020.07 | 55.2 | 37.6 | 30.3 | 68.9 | 51.5 | 43 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OC StereoTime=0.35s, Point cloud or pretrained disparity module=true2021.03 | 55.15 | 37.6 | 30.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Pseudo-LiDARTime=0.40s, Point cloud or pretrained disparity module=true2021.03 | 54.53 | 34.05 | 28.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PSEUDO-LIDAR2020.07 | 54.5 | 34.1 | 28.3 | 67.3 | 45 | 38.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| S-RCNN2020.07 | 47.6 | 30.2 | 23.7 | 61.9 | 41.3 | 33.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| StereoRCNNTime=0.30s, Point cloud or pretrained disparity module=false2021.03 | 47.58 | 30.23 | 23.72 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| F-POINTNETInput signal=Stereo, Representation=pseudo-LiDAR (PSMNet*), IoU threshold=0.72018.12 | 39.7 | 26.7 | 22.3 | 55 | 38.7 | 32.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DID-M3D+ours2026.02 | 30.69 | 20.45 | 17.25 | 39.68 | 26.5 | 22.69 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MonoDETR+ours2026.02 | 30.25 | 20.86 | 17.68 | 38.94 | 26.75 | 23.13 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MonoDGP+ours2026.02 | 30.25 | 21.1 | 17.85 | 37.95 | 26.82 | 23.11 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RT3DStereoTime=0.08s, Point cloud or pretrained disparity module=false2021.03 | 29.9 | 23.28 | 18.96 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| NeurOCSVenues=CVPR20232026.02 | 29.89 | 18.94 | 15.9 | 37.27 | 24.49 | 20.89 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MonoTAKDVenue=None, Extra Data=Raw2024.04 | 29.86 | 21.26 | 18.27 | 43.83 | 32.31 | 28.48 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GUPNet+ours2026.02 | 28.6 | 19.35 | 16.35 | 38.18 | 26.51 | 23.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CMKDVenue=ECCV 22, Extra Data=None2024.04 | 28.55 | 18.69 | 16.77 | 38.98 | 25.82 | 22.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — |