Visual Odometry on KITTI Seq. 10
0.62Translational Error (%)D3VO
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| D3VOTraining=Stereo, Geometric optimization=true2021.05 | 0.62 | — | — | — | — | |
| DVSOTraining=Stereo, Geometric optimization=true2021.05 | 0.74 | 0.21 | — | — | — | |
| Proposed-Adaptive2026.06 | 1.6 | 0.002 | — | 11.8292 | 10.2275 | |
| DF-VOSource=Original Paper, Protocol=Aligned using SE(3)2025.09 | 1.82 | 0.38 | 3 | — | — | |
| BEV-DWPVOProtocol=Aligned using SE(3)2025.09 | 3.24 | 1.15 | 9.04 | — | — | |
| ORB-SLAM22026.06 | 3.36 | 0.0028 | — | 5.4202 | — | |
| BEV-ODOM2Protocol=Aligned using SE(3)2025.09 | 3.46 | 0.52 | 6.08 | — | — | |
| BEV-ODOMProtocol=Aligned using SE(3)2025.09 | 3.61 | 0.53 | 8.42 | — | — | |
| Pseudo-RGBD SLAM (Ours - Motion Model)Training=Monocular, Geometric optimization=true2021.05 | 3.82 | 1.76 | 5.96 | — | — | |
| Pseudo-RGBD SLAM (Ours - Pose CNN)Training=Monocular, Geometric optimization=true2021.05 | 4.32 | 2.34 | 7.99 | — | — | |
| DPVOSource=Pretrained Model, Protocol=Scaled by the first 10m’s ground truth and aligned using SE(3)2025.09 | 4.47 | 0.23 | 10.98 | — | — | |
| Zhao et al.Training=Monocular, Geometric optimization=true2021.05 | 4.66 | 0.62 | — | — | — | |
| Zou et al.Training=Monocular2021.05 | 5.81 | 1.8 | 11.8 | — | — | |
| TartanVOSource=Original Paper, Protocol=Aligned using SE(3)2025.09 | 6.89 | 2.73 | 28.5 | — | — | |
| Manydepth22023.12 | 7.29 | 2.65 | — | — | — | |
| Pseudo-RGBD SLAM (Ours w/o LG)Training=Monocular, Geometric optimization=true2021.05 | 7.62 | 2.41 | 9.02 | — | — | |
| SC-Depth (Ours)Training=Monocular2021.05 | 7.79 | 4.9 | 12 | — | — | |
| ORB-SLAM3Protocol=Scaled by the first 10m’s ground truth and aligned using SE(3)2025.09 | 8.47 | 0.93 | 16.15 | — | — | |
| ManyDepth2023.12 | 9.86 | 3.42 | — | — | — | |
| UndeepVOTraining=Stereo2021.05 | 10.63 | 4.6 | — | — | — | |
| sfmLearner2026.06 | 10.68 | 0.0447 | — | 20.2817 | — | |
| MonoDepth2Training=Monocular2021.05 | 11.68 | 5.31 | 20.35 | — | — | |
| SC-Depth (Ours w/o LG)Training=Monocular2021.05 | 11.86 | 4.95 | 21.19 | — | — | |
| DeepMatchVOTraining=Monocular2021.05 | 12.18 | 5.9 | 24.44 | — | — | |
| DFR2023.12 | 12.45 | 3.46 | — | — | — | |
| Depth-VO-FeatTraining=Stereo2021.05 | 12.82 | 3.41 | 24.7 | — | — | |
| PoseGraphTraining=Stereo, Geometric optimization=true2021.05 | 12.9 | 3.17 | — | — | — | |
| NeuralBundler2023.12 | 12.9 | 3.71 | — | — | — | |
| DROID-SLAMSource=Pretrained Model, Protocol=Scaled by the first 10m’s ground truth and aligned using SE(3)2025.09 | 18.73 | 0.23 | 45.52 | — | — | |
| VISO22026.06 | 22.41 | 0.0305 | — | 47.8139 | — | |
| Depth-VO-Feat2026.06 | 22.754 | 0.027 | — | 12.9822 | — | |
| GeoNetTraining=Monocular2021.05 | 23.9 | 9 | 43.04 | — | — | |
| DeepVOProtocol=Aligned using SE(3)2025.09 | 30.46 | 14.42 | 180.1 | — | — | |
| SfMLearnerTraining=Monocular2021.05 | 40.4 | 17.69 | 67.34 | — | — | |
| GD-VIO2026.06 | 41.73 | 0.0562 | — | 24.293 | — | |
| BEV-DWPVOProtocol=Aligned using Sim(3)2025.09 | — | — | 7.8 | — | — | |
| BEV-ODOMProtocol=Aligned using Sim(3)2025.09 | — | — | 7.3 | — | — | |
| BEV-ODOM2Protocol=Aligned using Sim(3)2025.09 | — | — | 5.97 | — | — | |
| DeepVOProtocol=Aligned using Sim(3)2025.09 | — | — | 168.44 | — | — | |
| DF-VOSource=Original Paper, Protocol=Aligned using Sim(3)2025.09 | — | — | 2.73 | — | — | |
| DPVOSource=Pretrained Model, Protocol=Aligned using Sim(3)2025.09 | — | — | 10.99 | — | — | |
| DROID-SLAMSource=Pretrained Model, Protocol=Aligned using Sim(3)2025.09 | — | — | 17.27 | — | — | |
| DW-CorrectedTraining=Monocular2021.05 | — | — | 14.85 | — | — | |
| DW-LearnedTraining=Monocular2021.05 | — | — | 17.88 | — | — | |
| ORB-SLAM3Protocol=Aligned using Sim(3)2025.09 | — | — | 7.33 | — | — | |
| TartanVOSource=Original Paper, Protocol=Aligned using Sim(3)2025.09 | — | — | 23.65 | — | — |