Optical Flow on Middlebury (test)
0.32AEEClassic+NLP
Evaluation Results
| Method | Links | |
|---|---|---|
| Classic+NLPCategory=Conventional2018.05 | 0.32 | |
| Classic+NLPFine-tuned=false, Time (s)=1022016.11 | 0.32 | |
| FlowFieldsbackbone=CPU, runtime=22,810ms2016.12 | 0.33 | |
| EPPMbackbone=GPU, runtime=200ms2016.12 | 0.33 | |
| FlowFieldsCategory=Conventional2018.05 | 0.33 | |
| EpicFlow+backbone=CPU, runtime=42,600ms2016.12 | 0.39 | |
| EpicFlowCategory=Conventional2018.05 | 0.39 | |
| LiteFlowNet-ftCategory=Lightweight CNN, fine-tuned=true2018.05 | 0.4 | |
| DeepFlow+backbone=CPU, runtime=51,940ms2016.12 | 0.42 | |
| DeepFlowCategory=Conventional2018.05 | 0.42 | |
| FlowNet2backbone=GPU, runtime=123ms2016.12 | 0.52 | |
| FlowNet2Category=Heavyweight CNN2018.05 | 0.52 | |
| LDOF (CPU)backbone=CPU, runtime=64,900ms2016.12 | 0.56 | |
| LDOFCategory=Conventional2018.05 | 0.56 | |
| SPyNetCategory=Lightweight CNN2018.05 | 0.58 | |
| SPyNet-ftCategory=Lightweight CNN, fine-tuned=true2018.05 | 0.58 | |
| SPyNetFine-tuned=false, Time (s)=0.0692016.11 | 0.58 | |
| SPyNetFine-tuned=true, Time (s)=0.0692016.11 | 0.58 |