Scene Flow Estimation on FlyingThings3D with occlusions (F3Do) (test)
0.015EPE3DDiffSF
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DiffSFinference time steps=22024.03 | 0.015 | 96.7 | 98.1 | 3.5 | |
| GMSFTraining Dataset=F3Do2023.05 | 0.022 | 95 | 97.5 | 5.6 | |
| GMSFinference time steps=22024.03 | 0.022 | 95 | 97.5 | 5.6 | |
| DifFlow3Dinference time steps=22024.03 | 0.047 | 88.2 | 94 | 15 | |
| MSBRNinference time steps=22024.03 | 0.053 | 83.6 | 92.6 | 23.1 | |
| 3DFlowTraining Dataset=F3Do2023.05 | 0.063 | 79.1 | 90.9 | 27.9 | |
| 3DFlowinference time steps=22024.03 | 0.063 | 79.1 | 90.9 | 27.9 | |
| RAFT3DTraining Dataset=F3Do2023.05 | 0.064 | 83.7 | 89.2 | — | |
| Bi-PointFlowTraining Dataset=F3Do2023.05 | 0.073 | 79.1 | 89.6 | 27.4 | |
| Bi-PointFlowinference time steps=22024.03 | 0.073 | 79.1 | 89.6 | 27.4 | |
| CamLiRAFTTraining Dataset=F3Do2023.05 | 0.076 | 79.4 | 90.4 | 27.9 | |
| CamLiRAFT-LMethod Category=Recurrent, Iterative updates (Tx)=20x, Throughput (pair/s)=9.52023.03 | 0.076 | 79.4 | 90.4 | 27.9 | |
| CamLiRAFT-LMethod Category=Recurrent, Iterative updates (Tx)=8x, Throughput (pair/s)=20.82023.03 | 0.078 | 78.4 | 89.6 | 28.3 | |
| CamLiPWCTraining Dataset=F3Do2023.05 | 0.092 | 71.5 | 87.1 | 37.2 | |
| CamLiPWC-LMethod Category=Coarse-to-fine, Throughput (pair/s)=61.62023.03 | 0.092 | 71.5 | 87.1 | 37.2 | |
| OGSF-NetMethod Category=Coarse-to-fine, Throughput (pair/s)=14.12023.03 | 0.122 | 55.2 | 77.7 | 51.8 | |
| FLOTTraining Dataset=F3Do2023.05 | 0.153 | 39.6 | 66 | 66.2 | |
| FLOTMethod Category=Coarse-to-fine, Throughput (pair/s)=8.62023.03 | 0.153 | 39.7 | 66.1 | 66.3 | |
| FLOTinference time steps=22024.03 | 0.153 | 39.6 | 66 | 66.2 | |
| PointPWCTraining Dataset=F3Do2023.05 | 0.155 | 41.6 | 69.9 | 63.8 | |
| PointPWCMethod Category=Coarse-to-fine, Throughput (pair/s)=12.62023.03 | 0.155 | 41.6 | 69.9 | 63.9 | |
| PointPWCinference time steps=22024.03 | 0.155 | 41.6 | 69.9 | 63.8 | |
| FlowNet3DTraining Dataset=F3Do2023.05 | 0.157 | 22.8 | 58.2 | 80.4 | |
| FlowNet3Dinference time steps=22024.03 | 0.157 | 22.8 | 58.2 | 80.4 | |
| FlowNet3DMethod Category=Coarse-to-fine, Throughput (pair/s)=94.42023.03 | 0.158 | 22.9 | 58.2 | 80.4 | |
| HPLFlowNetTraining Dataset=F3Do2023.05 | 0.168 | 26.2 | 57.4 | 81.2 | |
| HPLFlowNetinference time steps=22024.03 | 0.168 | 26.2 | 57.4 | 81.2 | |
| HPLFlowNetMethod Category=Coarse-to-fine2023.03 | 0.169 | 26.3 | 57.5 | 81.2 |