Optical Flow Estimation on Su 2017 (Blurriest 10%)
26.64PSNR (dB)BIMNetFN2
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BIMNetFN2backbone=FlowNet2, loss=blur-invariant loss2021.08 | 26.64 | 0.873 | |
| BIMNetLFNbackbone=LiteFlowNet, loss=blur-invariant loss2021.08 | 26.03 | 0.853 | |
| FlowNet22021.08 | 25.79 | 0.864 | |
| TVL12021.08 | 25.08 | 0.784 | |
| FlowNet2sstraining_mode=sharp video frames (overfitting)2021.08 | 25.06 | 0.811 | |
| LiteFlowNetsstraining_mode=sharp video frames2021.08 | 24.89 | 0.83 | |
| LiteFlowNet2021.08 | 24.64 | 0.833 | |
| FlowNet2**training_mode=including blurry frames2021.08 | 24.14 | 0.761 | |
| LiteFlowNet**training_mode=including blurry frames2021.08 | 24.07 | 0.753 |