List your company

Optical Flow on KITTI (test)

1.8EPE (Overall)FlowNet2-ft-kitti
1.2964.6988.111.502Jun 25, 2015Nov 8, 2015Mar 23, 2016Aug 7, 2016Dec 21, 2016May 6, 2017Sep 20, 2017

Evaluation results

MethodLinks
FlowNet2-ft-kittibackbone=GPU, runtime=123ms, fine-tuned=KITTI2016.121.8————
BTF-ILLUMTime=80s2015.062.81.56.5211.03—
Flow FieldsTime=23s2015.063.51.46.2314.01—
FlowFieldsbackbone=CPU, runtime=22,810ms2016.123.5————
Discrete FlowTime=180s2015.063.61.35.7716.63—
EpicFlowTime=16s2015.063.81.57.8817.08—
NLTGV-SCTime=16s2015.063.81.65.9311.96—
EpicFlow+backbone=CPU, runtime=42,600ms2016.123.8————
EpicFlow2017.093.8————
TGV2ADCSIFTTime=12s2015.064.51.56.215.15—
SPyNetFine-tuned=true, Time (s)=0.0692016.114.7————
TF+OFMTime=350s2015.065210.2218.46—
PCA-Layersbackbone=GPU, runtime=3,300ms2016.125.2————
DeepFlowTime=22s2015.065.31.46.6117.35—
Data-FlowTime=180s2015.065.51.97.1114.57—
DeepFlow*Time=17s, Note=Preliminary version of DeepFlow2015.065.81.57.2217.79—
DeepFlow+backbone=CPU, runtime=51,940ms2016.125.8————
DeepFlow2017.095.8————
PCA-Flowbackbone=CPU, runtime=140ms2016.126.2————
SegFlow+ftfine-tuned=true2017.097.1————
FlowNetS+fttraining=reported in [12]2017.097.6————
FlowNetS+ft*training=same data as SegFlow2017.098.7————
FlowNetSFine-tuned=true, Time (s)=0.0802016.119.1————
EPPMbackbone=GPU, runtime=200ms2016.129.2————
EPPM2017.099.2————
LDOF (CPU)backbone=CPU, runtime=64,900ms2016.1212.4————
LDOF2017.0912.4————
DIS-Fastbackbone=CPU, runtime=70ms2016.1214.4————
CRAFTTraining dataset=Undisclosed, undisclosed method=true2022.04————4.79
DistractFlowwarm-start=false2024.03————4.71
FlowSupervisorwarm-start=false2024.03————4.85
GMATraining dataset=C+T+S+K+H2022.04————5.15
LiteFlowNet2Training dataset=C+T+S+K+H2022.04————7.62
MaskFlowNetTraining dataset=C+T+S+K+H2022.04————6.1
OCAIwarm-start=false2024.03————4.13
PWC-Net+Training dataset=C+T+S+K+H2022.04————7.72
RAFTTraining dataset=C+T+S+K+H2022.04————5.1
RAFT + OCTCTraining dataset=C+T+S+K+H2022.04————4.72
RAFT + OCTCTraining dataset=C+T+S+K+H, trained on KITTI-2012 and KITTI-2015 datasets=true2022.04————4.33
RAFT-ATraining dataset=A+T+S+K+H2022.04————4.78
RAFT-Awarm-start=false2024.03————4.78
RAFT-OCTCwarm-start=false2024.03————4.72
RAFT-Supervisedwarm-start=false2024.03————5.1
RealFlowwarm-start=false2024.03————4.63
ScopeFlowTraining dataset=C+T+S+K2022.04————6.82
SelFlowTraining dataset=C+T+S+K2022.04————8.42
VCNTraining dataset=C+T+S+K+H2022.04————6.3