List your company
ResearchFormerly WizWand

Optical Flow Estimation on MPI Sintel (test)

1EPE (Clean, all)SAMFlow
0.66482.92745.197.4526May 4, 2024
Updated 8mo ago

Evaluation results

MethodLinks
SAMFlowLearning Paradigm=Supervised, SAM used in training=true, SAM used in inference=true2024.0510.385.972.081.0410.6
FlowFormerLearning Paradigm=Supervised, Parameters=18.2M2024.051.160.427.162.090.9611.3
RAFTLearning Paradigm=Supervised, Parameters=5.3M2024.051.610.629.652.861.4114.68
PWC-Net+Learning Paradigm=Supervised, Parameters=8.8M2024.053.451.4120.124.62.2523.7
IRR-PWCLearning Paradigm=Supervised, Parameters=6.4M2024.053.841.4723.224.582.1524.36
UnSAMFlow (+aug +hg +mf)Learning Paradigm=Unsupervised, Parameters=2.6M, semantic augmentation module=true, homography smoothness loss=true, mask feature module=true, SAM used in training=true, SAM used in inference=true2024.053.931.6722.345.22.5626.75
UnSAMFlow (+aug)Learning Paradigm=Unsupervised, Parameters=2.5M, semantic augmentation module=true, SAM used in training=true2024.0541.5823.765.332.5328.17
UnSAMFlow (+aug +hg)Learning Paradigm=Unsupervised, Parameters=2.5M, semantic augmentation module=true, homography smoothness loss=true, SAM used in training=true2024.0541.7622.365.222.6226.4
UnSAMFlow (baseline)Learning Paradigm=Unsupervised, Parameters=2.5M2024.054.291.6425.965.812.7630.6
UPFlowLearning Paradigm=Unsupervised, Parameters=3.5M2024.054.681.7128.955.322.4228.93
ARFlowLearning Paradigm=Unsupervised, Parameters=2.2M2024.054.781.9128.265.892.7331.6
UFlowLearning Paradigm=Unsupervised2024.055.212.0431.066.53.0834.4
SimFlowLearning Paradigm=Unsupervised2024.055.932.1636.666.923.0238.7
DDFlowLearning Paradigm=Unsupervised, Parameters=4.3M2024.056.182.2738.057.43.4139.94
SelFlowLearning Paradigm=Unsupervised, Parameters=4.8M2024.056.562.6738.36.573.1234.72
UnFlow-CSSLearning Paradigm=Unsupervised, Parameters=116.6M2024.059.385.3742.1110.226.0644.11