Motion Estimation on nuScenes Speed <= 5m/s
0.2286Mean ErrorSelf-Supervised Pillar Motion Learning
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Self-Supervised Pillar Motion LearningTraining Protocol=Fine-tuned on nuScenes, Inference Time=0.020s2021.04 | 0.2286 | 0.093 | |
| MotionNet+MGDATraining Protocol=Supervised on nuScenes, Inference Time=0.019s2021.04 | 0.2292 | 0.0952 | |
| MotionNetTraining Protocol=Supervised on nuScenes, Inference Time=0.019s2021.04 | 0.2467 | 0.0961 | |
| MotionNet (pillar-based)Training Protocol=Supervised on nuScenes, Inference Time=0.019s2021.04 | 0.2612 | — | |
| LSTMEncoderDecoderTraining Protocol=Supervised on nuScenes, Inference Time=0.042s2021.04 | 0.3551 | 0.1044 | |
| HPLFlowNetTraining Protocol=Supervised on nuScenes, Inference Time=0.352s2021.04 | 0.4458 | 0.0969 | |
| PointRCNNTraining Protocol=Supervised on nuScenes, Inference Time=0.201s2021.04 | 0.5514 | 0.1627 | |
| Self-Supervised Pillar Motion LearningTraining Protocol=Self-supervised, Inference Time=0.020s2021.04 | 0.6972 | 0.1758 | |
| FlowNet3DTraining Protocol=Supervised on nuScenes, Inference Time=0.434s2021.04 | 0.8183 | 0.1782 | |
| HPLFlowNetTraining Protocol=Pre-trained on FlyingThings3D and KITTI, Inference Time=0.352s2021.04 | 1.5477 | 1.1269 | |
| FlowNet3DTraining Protocol=Pre-trained on FlyingThings3D and KITTI, Inference Time=0.434s2021.04 | 2.2058 | 0.3172 |