Scene Flow Estimation on Argoverse 2
EPE 3-Way
- 2021-03-01 — FastFlow3D: EPE 3-Way 0.06196
- 2023-04-18 — FastNSF: EPE 3-Way 0.11182
| Rank | Model | EPE 3-Way | EPE Foreground Dynamic | EPE Foreground Static | EPE Background Static | Extra Training Data | Paper | Code | Year |
|---|---|---|---|---|---|---|---|---|---|
| 1 | DeFlow | 0.034295 | 0.073231 | 0.025093 | 0.004561 | DeFlow: Decoder of Scene Flow Network in Autonomous Driving | kth-rpl/dynamicmap_benchmark · kth-rpl/deflow · kth-rpl/seflow · +1 | 2024 | |
| 2 | TrackFlow | 0.047316 | 0.102975 | 0.036544 | 0.002428 | ||||
| 3 | SeFlow | 0.048590 | 0.121375 | 0.018404 | 0.005990 | SeFlow: A Self-Supervised Scene Flow Method in Autonomous Driving | kth-rpl/deflow · kth-rpl/seflow | 2024 | |
| 4 | ZeroFlow 5x XL | 0.049392 | 0.117688 | 0.017406 | 0.013082 | ✓ | ZeroFlow: Scalable Scene Flow via Distillation | kylevedder/zeroflow | 2023 |
| 5 | NSFP | 0.060606 | 0.115796 | 0.031576 | 0.034445 | Neural Scene Flow Prior | lilac-lee/neural_scene_flow_prior | 2021 | |
| 6 | FastFlow3D | 0.061960 | 0.156392 | 0.024549 | 0.004939 | Scalable Scene Flow from Point Clouds in the Real World | kth-rpl/deflow · leolyj/dca-srsfe · jabb0/fastflow3d · +2 | 2021 | |
| 7 | FastNSF | 0.111820 | 0.163388 | 0.081360 | 0.090712 | Fast Neural Scene Flow | lilac-lee/fastnsf | 2023 |