3D Multi-Object Tracking on nuScenes (test)
93ID SwitchesNEBP
Evaluation Results
| Method | Links | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NEBPDetections=CenterPoint [26]2023.10 | 93 | — | 0.598 | 61.9 | — | — | — | — | 74.1 | 70.8 | 83.5 | 70.8 | 59.8 | 69 | — | — | — | — | |
| Immortal TrackerDetections=CenterPoint [26]2023.10 | 155 | — | 0.609 | 59.9 | — | — | — | — | 74.5 | 70.5 | 83.3 | 71.6 | 59.6 | 67.5 | — | — | — | — | |
| ShaSTADetections=CenterPoint [26]2023.10 | 185 | — | 0.559 | 61.2 | — | — | — | — | 74.3 | 73.1 | 83.8 | 73.3 | 65 | 70.4 | — | — | — | — | |
| Fast-Poly-BDetector(s)=MV2DFusion [46], Input=Structure, Offline=true2026.02 | 203 | 76.7 | 0.489 | 64.9 | — | 19,342 | 15,972 | — | — | — | — | — | — | — | — | — | — | — | |
| Offline Re-ID (Motion + Map)Detections=CenterPoint [26]2023.10 | 204 | — | 0.532 | 61.3 | — | — | — | — | 74.2 | 73.4 | 84.2 | 75.1 | 64.1 | 70.3 | — | — | — | — | |
| Offline-PolyDetector(s)=MV-FP-F & MV-FP-B, Input=Structure, Offline=true2026.02 | 208 | 77.6 | 0.486 | 65.5 | — | 19,098 | 15,929 | — | — | — | — | — | — | — | — | — | — | — | |
| 3DMOTFormer++Detections=CenterPoint [26]2023.10 | 210 | — | 0.542 | 58.6 | — | — | — | — | 73 | 72.3 | 82.1 | 74.9 | 62.6 | 69.6 | — | — | — | — | |
| MCTrackDetector=LargeKernel3D [11]2024.09 | 242 | — | — | — | 103,327 | 19,643 | 15,996 | — | — | 76.3 | 87.9 | 77.2 | 66.7 | 77.9 | — | 59.1 | 81.9 | 83.2 | |
| MCTrackDetector(s)=LargeKernel-F [61], Input=Structure, Offline=false2026.02 | 242 | 76.3 | 0.445 | 63.4 | — | 19,643 | 15,996 | — | — | — | — | — | — | — | — | — | — | — | |
| MC-TrackDetector=LK-F, Input=C + L2026.03 | 242 | 76.3 | — | 63.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PF-TrackResolution=Full (F)2023.02 | 249 | 0.434 | 1.252 | 0.378 | — | — | — | — | 53.8 | — | — | — | — | — | — | — | — | — | |
| PF-TrackE2E=true, Detector=DETR3D, Backbone=V2-992025.12 | 249 | 43.4 | 1.252 | 37.8 | — | 19,048 | 42,758 | — | — | — | — | — | — | — | — | — | — | — | |
| SimpleTrackDetections=CenterPoint [26]2023.10 | 259 | — | 0.582 | 58.6 | — | — | — | — | 73.7 | 70 | 82.3 | 71.5 | 58.7 | 67.3 | — | — | — | — | |
| Fast-Poly-FDetector(s)=MV2DFusion [46], Input=Structure, Offline=false2026.02 | 266 | 76.3 | 0.505 | 64.5 | — | 20,083 | 15,812 | — | — | — | — | — | — | — | — | — | — | — | |
| RegTrackInput=Camera and LiDAR, 3D Object Detector=LargeKernel3D2024.09 | 266 | — | — | — | 101,980 | 16,611 | 17,319 | — | 77.8 | 76.5 | 87.2 | 77.6 | 66 | 78.6 | — | 58.5 | 81.9 | 85.7 | |
| Easy-PolyDevice=4090‡, Detector=LargeKernel3D [6], Input=2D+3D2025.02 | 270 | 76.3 | — | 63.5 | — | 16,319 | 16,602 | — | — | — | — | — | — | — | 34.8 | — | — | — | |
| EMMS-MOTDetector=LK-F & CR, Input=C + L2026.03 | 271 | 76.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Fusion-PolyDetector=MV2D & CR, Input=C + L2026.03 | 282 | 76.5 | — | 65.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OGR3MOTDetection Source=CenterPoint [43]2021.11 | 288 | 0.656 | 0.62 | 0.554 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OGR3MOTModality=Lidar2021.11 | 288 | 65.6 | — | 55.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OGR3MOTInput modality=LiDAR, Double-flip testing=false2023.03 | 288 | 65.6 | 62 | 55.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OGR3MOTModality=LiDAR2023.08 | 288 | 65.6 | 62 | 55.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Poly-MOTDevice=9940X, Detector=LargeKernel3D [25], Input=2D+3D2024.03 | 292 | 75.4 | — | 62.1 | — | 19,673 | 17,956 | — | — | — | — | — | — | — | 3 | — | — | — | |
| Poly-MOTDetector=LargeKernel3D [11]2024.09 | 292 | — | — | — | 101,317 | 19,673 | 17,956 | — | — | 75.4 | 86.5 | 78.6 | 66.2 | 75.1 | — | 58.2 | 81 | 82 | |
| Poly-MOTDevice=9940X, Detector=LargeKernel3D [6], Input=2D+3D2025.02 | 292 | 75.4 | — | 62.1 | — | 19,673 | 17,956 | — | — | — | — | — | — | — | 3 | — | — | — | |
| Poly-MOTDetector(s)=LargeKernel-F [61], Input=Structure, Offline=false2026.02 | 292 | 75.4 | 0.422 | 62.1 | — | 17,956 | 19,673 | — | — | — | — | — | — | — | — | — | — | — | |
| Poly-MOTInput=LiDAR, 3D Object Detector=LargeKernel3D2024.09 | 292 | — | — | — | 101,317 | 19,673 | 17,956 | — | 78.3 | 75.4 | 86.5 | 78.6 | 66.2 | 75.1 | — | 58.2 | 81 | 82 | |
| Poly-MOTDetector=LK-F, Input=C + L2026.03 | 292 | 75.4 | — | 62.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RegTrackInput=Camera and LiDAR, 3D Object Detector=FocalFormer3D-F2024.09 | 319 | — | — | — | 103,868 | 15,827 | 15,378 | — | 79.9 | 78 | 88 | 78.7 | 68.3 | 77.6 | — | 63.1 | 83.2 | 86.8 | |
| CAMO-MOTDetector=BEVFusion [34] & FC [10]2024.09 | 324 | — | — | — | 101,049 | 17,269 | 18,192 | — | — | 75.3 | 85.8 | 77.7 | 67.7 | 72.3 | — | 59.2 | 78.2 | 85.8 | |
| CAMO-MOTDetector(s)=BEVFusion [59] & FocalsConv [60], Input=Structure & Semantic, Offline=false2026.02 | 324 | 75.3 | 0.472 | 63.5 | — | 17,269 | 18,192 | — | — | — | — | — | — | — | — | — | — | — | |
| CAMO-MOTInput=Camera and LiDAR2024.09 | 324 | — | — | — | 101,049 | 17,269 | 18,192 | — | 79.1 | 75.3 | 85.8 | 77.7 | 67.7 | 72.3 | — | 59.2 | 78.2 | 85.8 | |
| CAMO-MOTDetector=BEVF & FC, Input=C + L2026.03 | 324 | 75.3 | — | 63.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MinkowskiDevice=TITAN, Detector=Minkowski [5], Input=3D2024.03 | 325 | 69.8 | — | 57.8 | — | 19,340 | 21,200 | — | — | — | — | — | — | — | 3.5 | — | — | — | |
| MinkowskiDevice=TITAN‡, Detector=Minkowski [9], Input=3D2025.02 | 325 | 69.8 | — | 57.8 | — | 19,340 | 21,200 | — | — | — | — | — | — | — | 3.5 | — | — | — | |
| Fast-PolyDevice=7945HX, Detector=LargeKernel3D [25], Input=2D+3D2024.03 | 326 | 75.8 | — | 62.8 | — | 17,098 | 18,415 | — | — | — | — | — | — | — | 34.2 | — | — | — | |
| Fast-PolyDetector=LargeKernel3D [11]2024.09 | 326 | — | — | — | 100,824 | 17,098 | 18,415 | — | — | 75.8 | 86.3 | 76.7 | 65.6 | 76.8 | — | 57.3 | 82.6 | 85.2 | |
| Fast-PolyDevice=7945HX, Detector=LargeKernel3D [6], Input=2D+3D2025.02 | 326 | 75.8 | — | 62.8 | — | 17,098 | 18,415 | — | — | — | — | — | — | — | 34.2 | — | — | — | |
| Fast-PolyDetector(s)=LargeKernel-F [61], Input=Structure, Offline=false2026.02 | 326 | 75.8 | 0.479 | 62.8 | — | 17,098 | 18,415 | — | — | — | — | — | — | — | — | — | — | — | |
| Fast-PolyInput=LiDAR, 3D Object Detector=LargeKernel3D2024.09 | 326 | — | — | — | 100,824 | 17,098 | 18,415 | — | 77.6 | 75.8 | 86.3 | 76.7 | 65.6 | 76.8 | — | 57.3 | 82.6 | 85.2 | |
| Fast-PolyDetector=LK-F, Input=C + L2026.03 | 326 | 75.8 | — | 62.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CenterPointDetections=CenterPoint [26]2023.10 | 340 | — | 0.596 | 59.4 | — | — | — | — | 73.5 | 69.8 | 82.9 | 71.1 | 59.9 | 65.1 | — | — | — | — | |
| Immortal TrackerModality=Lidar2021.11 | 365 | 66.1 | — | 55.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| NEBP-V3Detector=CenterPoint [68] & FC [10]2024.09 | 377 | — | — | — | 99,872 | 17,243 | 19,316 | — | — | 74.6 | 86.1 | 78.6 | 67.3 | 76.1 | — | 49.9 | 80.7 | 83.4 | |
| NEBP-V3Input=LiDAR2024.09 | 377 | — | — | — | 99,872 | 17,243 | 19,316 | — | 76.3 | 74.6 | 86.1 | 78.6 | 67.3 | 76.1 | — | 49.9 | 80.7 | 83.4 | |
| UVTRDetections=CenterPoint [26]2023.10 | 381 | — | 0.636 | 59.3 | — | — | — | — | 74.6 | 70.1 | 83.3 | 67.2 | 58.4 | 71.6 | — | — | — | — | |
| DINO-MOTDetector(s)=Is-Fusion [62], Input=Structure & Semantic, Offline=false2026.02 | 387 | 76.3 | 0.52 | 65 | — | 15,956 | 17,070 | — | — | — | — | — | — | — | — | — | — | — | |
| DINO-MOTDetector=-, Input=C + L2026.03 | 387 | 76.3 | — | 65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DualViewDistillInput=Camera2024.09 | 407 | — | — | — | 96,276 | 16,499 | 22,882 | — | 74.1 | 66.9 | 78.8 | 63.3 | 56.8 | 67 | — | 57.3 | 70.6 | 74.3 | |
| 3DMOTFormerDetector=CenterPoint [68]2024.09 | 438 | — | — | — | 95,790 | 18,322 | 23,337 | — | — | 68.2 | 82.1 | 74.9 | 62.6 | 69.6 | — | 37.4 | 70.5 | 80.7 | |
| 3DMOTFormerInput=LiDAR2024.09 | 438 | — | — | — | 95,790 | 18,322 | 23,337 | — | 69.3 | 68.2 | 82.1 | 74.9 | 62.6 | 69.6 | — | 37.4 | 70.5 | 80.7 | |
| FutrTrackInput=Camera and LiDAR, 3D Object Detector=FocalFormer3D-F2024.09 | 447 | — | — | — | 99,204 | 18,295 | 19,914 | — | 78 | 74.7 | 83.8 | 79 | 67.8 | 75.4 | — | 55.8 | 76.8 | 84.2 | |
| ShaSTADetector=CenterPoint [68]2024.09 | 473 | — | — | — | 97,799 | 16,746 | 21,293 | — | — | 69.6 | 83.8 | 73.3 | 65 | 70.4 | — | 41 | 72.7 | 81 | |
| ShaSTAInput=LiDAR2024.09 | 473 | — | — | — | 97,799 | 16,746 | 21,293 | — | 71 | 69.6 | 83.8 | 73.3 | 65 | 70.4 | — | 41 | 72.7 | 81 | |
| ByteTrackv2Device=TITAN, Detector=TransFusion-L [35], Input=3D2024.03 | 488 | 70.1 | — | 58 | — | 18,682 | 21,836 | — | — | — | — | — | — | — | — | — | — | — | |
| ByteTrackv2Detector=TransFusion-L [1], Input=3D2025.02 | 488 | 70.1 | — | 58 | — | 18,682 | 21,836 | — | — | — | — | — | — | — | — | — | — | — | |
| BEVFusion-MITModality=Multi-Modal, Model ensembling=true2023.08 | 506 | 74.1 | 40.3 | 60.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BEVFusionDetector=BEVFusion [34]2024.09 | 506 | — | — | — | 99,664 | 19,997 | 19,395 | — | — | 74.1 | 83.1 | 77.9 | 69.5 | 73.4 | — | 56 | 74.8 | 83.7 | |
| BEVFusionInput=Camera and LiDAR2024.09 | 506 | — | — | — | 99,664 | 19,997 | 19,395 | — | 77.9 | 74.1 | 83.1 | 77.9 | 69.5 | 73.4 | — | 56 | 74.8 | 83.7 | |
| CBMOTDetection Source=CenterPoint [43]2021.11 | 557 | 0.649 | 0.592 | 0.545 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CBMOTModality=Lidar2021.11 | 557 | 64.9 | — | 54.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CBMOTInput modality=LiDAR, Double-flip testing=false2023.03 | 557 | 64.9 | 59.2 | 54.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CBMOTModality=LiDAR2023.08 | 557 | 64.9 | 59.2 | 54.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OTOPDetector(s)=CenterPoint [45], Input=Structure, Offline=true2026.02 | 570 | 67.1 | 0.522 | 55.3 | — | 16,778 | 22,378 | — | — | — | — | — | — | — | — | — | — | — | |
| SimpleTrackDetection Source=CenterPoint [43], Frame Rate=10Hz2021.11 | 575 | 0.668 | 0.55 | 0.566 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimpleTrackInput modality=LiDAR, Double-flip testing=false2023.03 | 575 | 66.8 | 55 | 56.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimpleTrackModality=LiDAR2023.08 | 575 | 66.8 | 55 | 56.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimpleTrackDetector=CenterPoint [68]2024.09 | 575 | — | — | — | 95,539 | 17,514 | 24,351 | — | — | 66.8 | 82.3 | 71.5 | 58.7 | 67.3 | — | 40.7 | 67.4 | 79.6 | |
| SimpleTrackInput=LiDAR2024.09 | 575 | — | — | — | 95,539 | 17,514 | 24,351 | — | 70.3 | 66.8 | 82.3 | 71.5 | 58.7 | 67.3 | — | 40.7 | 67.4 | 79.6 | |
| 3DMOTFormerDevice=2080Ti‡, Detector=BEVFuison [24], Input=2D+3D2024.03 | 593 | 72.5 | — | 60.9 | — | 17,530 | 20,996 | — | — | — | — | — | — | — | 54.7 | — | — | — | |
| 3DMOTFormerDevice=2080Ti‡, Detector=BEVFuison [20], Input=2D+3D2025.02 | 593 | 72.5 | — | 60.9 | — | 17,530 | 20,996 | — | — | — | — | — | — | — | 54.7 | — | — | — | |
| STAR-TrackE2E=true, Detector=DETR3D, Backbone=V2-992025.12 | 607 | 43.9 | 1.256 | 40.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimpleTrackDetection Source=CenterPoint [43], Frame Rate=2Hz2021.11 | 609 | 0.658 | 0.568 | 0.557 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CenterPoint++Modality=Lidar2021.11 | 609 | 65.8 | — | 55.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoxelNeXtInput modality=LiDAR, Double-flip testing=true2023.03 | 654 | 71 | 51.1 | 60 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VoxelNeXtDetector=VoxelNeXt [12]2024.09 | 654 | — | — | — | 97,075 | 18,348 | 21,836 | — | — | 71 | 82.6 | 74.7 | 64.4 | 73.8 | — | 52.6 | 73.1 | 76 | |
| VoxelNeXtInput=LiDAR2024.09 | 654 | — | — | — | 97,075 | 18,348 | 21,836 | — | 76.5 | 71 | 82.6 | 74.7 | 64.4 | 73.8 | — | 52.6 | 73.1 | 76 | |
| FocalFormer3DModality=LiDAR, Double-flip testing=true2023.08 | 701 | 72.1 | 47 | 60 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ByteTrackV2Input=C, Backbone=V2-992024.03 | 704 | 56.4 | 1.005 | — | — | 18,939 | 33,531 | — | — | — | — | — | — | — | — | — | — | — | |
| CBMOTDevice=i7-9700, Detector=CenterPoint [19]&CenterTrack [20], Input=2D+3D2024.03 | 709 | 67.6 | — | 53.9 | — | 21,604 | 22,828 | — | — | — | — | — | — | — | 80.5 | — | — | — | |
| CBMOTDevice=I7-9700, Detector=CenterPoint [32]&CenterTrack [35], Input=2D+3D2025.02 | 709 | 67.6 | — | 53.9 | — | 21,604 | 22,828 | — | — | — | — | — | — | — | 80.5 | — | — | — | |
| AlphaTrack2022.03 | 718 | 69.3 | — | — | 95,851 | 18,421 | 22,996 | — | — | — | — | — | — | — | — | — | — | — | |
| CenterPointDetection Source=CenterPoint [43]2021.11 | 760 | 0.638 | 0.555 | 0.537 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CenterPointModality=Lidar2021.11 | 760 | 63.8 | — | 53.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CenterPoint2022.03 | 760 | 63.8 | — | — | 95,877 | 18,612 | 22,928 | — | — | — | — | — | — | — | — | — | — | — | |
| CenterPointInput modality=LiDAR, Double-flip testing=false2023.03 | 760 | 63.8 | 55.5 | 53.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CenterPointModality=LiDAR2023.08 | 760 | 63.8 | 55.5 | 53.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CenterPointInput=L, Backbone=Voxel2024.03 | 760 | 63.8 | 0.555 | — | — | 18,612 | 22,928 | — | — | — | — | — | — | — | — | — | — | — | |
| UVTR-LInput modality=LiDAR, Double-flip testing=false2023.03 | 774 | 67 | 55 | 56.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UVTR-LModality=LiDAR2023.08 | 774 | 67 | 55 | 56.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Chiu et al.Detection Source=CenterPoint [43]2021.11 | 776 | 0.55 | 0.798 | 0.459 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| StreamPETRInput=C, Backbone=V2-992024.03 | 784 | 56.6 | 0.975 | — | — | 21,268 | 31,484 | — | — | — | — | — | — | — | — | — | — | — | |
| VoxelNeXtInput modality=LiDAR, Double-flip testing=false2023.03 | 785 | 69.5 | 56.8 | 58.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HyDRaInput=C+R, Backbone=V2-992024.03 | 791 | 58.4 | 0.95 | — | — | 13,996 | 32,950 | — | — | — | — | — | — | — | — | — | — | — | |
| FocalFormer3D-FModality=Multi-Modal, Double-flip testing=false2023.08 | 824 | 73.9 | 51.4 | 61.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FocalFormer3D-FDetector=FocalFormer3D [13]2024.09 | 824 | — | — | — | 97,987 | 15,547 | 20,754 | — | — | 73.9 | 83.4 | 79.2 | 66.9 | 75.2 | — | 54.1 | 74.6 | 84.1 | |
| FocalFormer3D-FInput=Camera and LiDAR, 3D Object Detector=FocalFormer3D-F2024.09 | 824 | — | — | — | 97,987 | 15,547 | 20,754 | — | 75.9 | 73.9 | 83.4 | 79.2 | 66.9 | 75.2 | — | 54.1 | 74.6 | 84.1 | |
| ADA-TrackDetector=DETR3D [60]2024.09 | 834 | — | — | — | 79,051 | 15,699 | 39,680 | — | — | 45.6 | 66.4 | 38.2 | 35.9 | 43.7 | — | 33.4 | 48.4 | 53.4 | |
| ADA-TrackE2E=true, Detector=DETR3D, Backbone=V2-992025.12 | 834 | 45.6 | 1.237 | 40.6 | — | 15,699 | 39,680 | — | — | — | — | — | — | — | — | — | — | — | |
| FocalFormer3D-FModality=Multi-Modal, Double-flip testing=true2023.08 | 849 | 74.6 | 47.3 | 63 | — | — | — | — | — | — | — | — | — | — | — | — | — | — |