List your company
ResearchFormerly WizWand

Optical Flow on Sintel (train)

0.28AEPE (Clean)StreamFlow
0.24040.50770.7751.0423Apr 6, 2021Feb 2, 2022Dec 2, 2022Sep 30, 2023Jul 29, 2024May 27, 2025Mar 26, 2026
Updated 7mo ago

Evaluation results

MethodLinks
StreamFlowTraining Data=(C+)T+S+K+H, Warm-start strategy=false2023.110.280.38
VideoFlow-BOFTraining Data=(C+)T+S+K+H, Warm-start strategy=false2023.110.370.54
FlowFormer++Training Data=C+T+S+K+H2023.030.40.6
TransFlowTraining Data=(C+)T+S+K+H, Warm-start strategy=false2023.110.420.69
FlowFormerTraining Data=C+T+S+K+H2022.030.480.74
FlowFormerTraining Data=C+T+S+K+H2023.030.480.74
FlowFormerTraining Data=(C+)T+S+K+H, Warm-start strategy=false2023.110.480.74
SKFlowTraining Data=C+T+S+K+H, warm-start strategy=true2023.030.520.78
SKFlowTraining Data=(C+)T+S+K+H, Warm-start strategy=true2023.110.520.78
MFCFlowTraining Data=(C+)T+S+K+H, Warm-start strategy=false2023.110.560.89
GMFlowNetTraining Data=C+T+S+K+H2023.030.590.91
GMFlowNetTraining Data=(C+)T+S+K+H, Warm-start strategy=false2023.110.590.91
CRAFTTraining Data=C + T + S/K + H2022.030.61.06
CRAFTTraining Data=C+T+S+K+H2023.030.61.06
RFPMTraining Data=C + T + S/K + H2022.030.611.05
GMATraining Data=C + T + S/K (+ H), Condition=Fine-tuned2021.040.621.06
GMATraining Data=C + T + S/K + H2022.030.621.06
GMATraining dataset=C+T+S+K+H2022.040.621.06
GMATraining Data=C+T+S+K+H, warm-start strategy=true2022.030.621.06
GMATraining Data=C+T+S+K+H, warm-start strategy=true2023.030.621.06
GMATraining Data=(C+)T+S+K+H, Warm-start strategy=true2023.110.621.06
GMA (position-only)Training Data=C + T + S/K (+ H), Condition=Fine-tuned2021.040.641.08
FlowFormerTraining=C+T, Params (M)=18.22023.030.641.5
MFRFlowTraining Data=(C+)T+S+K+H, Warm-start strategy=false2023.110.641.04
GMA (joint position+content)Training Data=C + T + S/K (+ H), Condition=Fine-tuned2021.040.651.11
AGFlowTraining Data=(C+)T+S+K+H, Warm-start strategy=true2023.110.651.07
Separable FlowTraining Data=C + T + S/K + H2022.030.691.1
S-FlowTraining Data=C+T+S+K+H2022.030.691.1
S-FlowTraining Data=C+T+S+K+H2023.030.691.1
Separable FlowTraining Data=(C+)T+S+K+H, Warm-start strategy=false2023.110.691.1
GeoViTTraining Data=C + T, tiling technique=true2025.120.691.78
RAFT + OCTCTraining dataset=C+T+S+K+H2022.040.731.23
RAFT + OCTCTraining dataset=C+T+S+K+H, trained including test images as unlabeled data=true2022.040.741.24
RAFTTraining Data=C + T + S/K (+ H), Condition=Fine-tuned2021.040.761.22
RAFTTraining Data=C + T + S/K + H2022.030.761.22
RAFTTraining dataset=C+T+S+K+H2022.040.761.22
RAFTTraining Data=C+T+S+K+H2022.030.761.22
RAFTTraining Data=C+T+S+K+H, warm-start strategy=false2023.030.761.22
RAFTTraining Data=(C+)T+S+K+H, Warm-start strategy=true2023.110.761.22
RAFTTraining Data=C+T+S+K+H, warm-start strategy=true2022.030.771.27
RAFTTraining Data=C+T+S+K+H, warm-start strategy=true2023.030.771.27
FM-RAFTTraining Data=C+T+S+K+H2022.030.791.7
FM-RAFTTraining Data=C+T+S+K+H2023.030.791.7
MemFlow-TFrames=3, Zero-shot=true2026.030.852.06
MegaFlowFrames=4, Zero-shot=true2026.030.851.83
FlowDiffuserFrames=2, Zero-shot=true2026.030.862.19
StreamFlowTraining Data=(C+)T, Warm-start strategy=false2023.110.872.11
SAMFlowTraining Data=C + T, tiling technique=false2025.120.872.11
SAMFlowFrames=2, Zero-shot=true2026.030.872.11
StreamFlowFrames=4, Zero-shot=true2026.030.872.11
EMD-LSource=ICCV232023.040.882.55
MegaFlowFrames=2, Zero-shot=true2026.030.892.07
FlowFormer++Training Data=C+T2023.030.92.3
FlowFormer++Training Data=C + T, tiling technique=true2025.120.92.3
Flowformer++Frames=2, Zero-shot=true2026.030.92.3
TransFlowTraining Data=(C+)T, Warm-start strategy=false2023.110.932.33
TransFlowSource=CVPR232023.040.932.33
MemFlowFrames=3, Zero-shot=true2026.030.932.08
FlowFormerTraining Data=C+T2023.030.942.33
FlowFormer++Source=CVPR232023.040.942.33
FlowFormerTraining Data=C + T, tiling technique=true2025.120.942.33
FlowFormerTraining Data=(C+)T, Warm-start strategy=false2023.1112.45
FlowFormerTraining Data=C+T2022.031.012.4
FlowFormerFrames=2, Zero-shot=true2026.031.012.4
GAFlowTraining Data=(C+)T, Warm-start strategy=false2023.111.022.45
DPFlowTraining Data=C + T, tiling technique=false2025.121.022.26
VideoFlow-BOFTraining Data=(C+)T, Warm-start strategy=false2023.111.032.19
VideoFlow-BOF*Frames=3, Zero-shot=true2026.031.032.19
DropRAFTSource=Ours2023.041.062.25
GMFlowTraining Data=C+T2023.031.082.48
GMFlowTraining=C+T, Params (M)=4.72023.031.082.48
GMFlowSource=ICCV212023.041.082.48
GMFlowTraining Data=C + T, tiling technique=false2025.121.082.48
GMFlowFrames=2, Zero-shot=true2026.031.082.48
AnyFlow (R.E.)Training=C+T, Params (M)=5.42023.031.12.52
AnyFlowFrames=2, Zero-shot=true2026.031.12.52
RPKNetSource=AAAI242023.041.122.45
RPKNetFrames=2, Zero-shot=true2026.031.122.45
GMFlowNetTraining Data=C+T2023.031.142.71
GMFlowNetTraining=C+T, Params (M)=9.32023.031.142.71
GMFlowNetSource=CVPR222023.041.142.71
GMFlowNetTraining Data=C + T, tiling technique=false2025.121.142.71
UFMFrames=2, Zero-shot=true2026.031.152.01
AnyFlow + GMATraining=C+T, Params (M)=6.02023.031.162.62
AnyFlow (dynamic)Training=C+T, Params (M)=5.42023.031.172.58
VideoFlow-MOF*Frames=5, Zero-shot=true2026.031.182.56
SEA-RAFT (L)Source=ECCV242023.041.194.11
SEA-RAFT (L)Frames=2, Zero-shot=true2026.031.194.11
FlowFormer (small)Training=C+T, Params (M)=6.22023.031.22.64
SEA-RAFT (M)Source=ECCV242023.041.214.04
SKFlowTraining Data=C+T2023.031.222.46
SKFlowTraining=C+T, Params (M)=6.32023.031.222.46
SKFlowTraining Data=(C+)T, Warm-start strategy=false2023.111.222.46
SKFlowSource=NeurIPS222023.041.222.46
SKFlowTraining Data=C + T, tiling technique=false2025.121.222.46
CRAFTTraining Data=C + T / Autoflow2022.031.272.79
CRAFTTraining Data=C+T2023.031.272.79
CRAFTTraining=C+T, Params (M)=6.32023.031.272.79
CRAFTTraining Data=(C+)T, Warm-start strategy=false2023.111.272.79
CRAFTSource=CVPR222023.041.272.79