List your company
ResearchFormerly WizWand

Video Recognition on Kinetics 400 1.0 (val)

84.9Top-1 AccVideo Swin-L (384↑)
71.48474.96778.4581.933May 25, 2021Jul 28, 2021Sep 30, 2021Dec 3, 2021Feb 5, 2022Apr 10, 2022Jun 13, 2022
Updated 8mo ago

Evaluation results

MethodLinks
Video Swin-L (384↑)Pre-train=IN-21K, GFLOPs=2107, Views=10×5, Params=200.02022.0684.996.7
Video Swin-BPre-train=IN-21K, GFLOPs=282, Views=4 × 3, Params=88.12022.0682.795.5
MLP-3D-LPre-train=IN-1K, GFLOPs=308, Views=4 × 3, Params=135.62022.0681.495.2
ViViT-L/16 × 2 320Pre-train=IN-21K, GFLOPs=3992, Views=4 × 3, Params=310.82022.0681.394.7
MViT-B, 64 x 3Pre-train=None, GFLOPs=455, Views=3 × 3, Params=36.62022.0681.295.1
MLP-3D-MPre-train=IN-1K, GFLOPs=170, Views=4 × 3, Params=80.52022.068194.9
TimeSformer-LPre-train=IN-21K, GFLOPs=2380, Views=1 × 3, Params=121.42022.0680.794.7
Video Swin-SPre-train=IN-1K, GFLOPs=166, Views=4 × 3, Params=49.82022.0680.694.5
Video Swin-BPre-train=IN-1K, GFLOPs=282, Views=4 × 3, Params=88.12022.0680.694.6
ViViT-L/16 × 2Pre-train=IN-21K, GFLOPs=1446, Views=4 × 3, Params=310.82022.0680.694.7
DG-P3DPre-train=IN-1K, GFLOPs=218, Views=10×32022.0680.594.6
X3D-XXLPre-train=None, GFLOPs=144, Views=10×3, Params=20.32022.0680.494.6
TokShiftPre-train=IN-21K, GFLOPs=2096, Views=10×3, Params=303.42022.0680.494.4
MViT-B, 32 × 3Pre-train=None, GFLOPs=170, Views=1 × 5, Params=36.62022.0680.294.4
MLP-3D-SPre-train=IN-1K, GFLOPs=102, Views=4 × 3, Params=68.52022.0680.293.8
CorrNet-101Pre-train=None, GFLOPs=224, Views=10×32022.0679.893.9
SlowFast R101+NLPre-train=None, GFLOPs=234, Views=10×3, Params=59.92022.0679.893.9
LGD-3DPre-train=IN-1K, GFLOPs=1952022.0679.494.4
TDNPre-train=IN-1K, GFLOPs=198, Views=10×32022.0679.494.4
DSA+I3DBackbone=3D ResNet-50, Sampling (Tinfer x Uinfer x #crop)=(4+8)x8x3, GFLOPs=251.5x2x3=15092021.057993.7
Video Swin-TPre-train=IN-1K, GFLOPs=88, Views=4 × 3, Params=28.22022.0678.893.6
SmallBigPre-train=IN-1K, GFLOPs=475, Views=4 × 32022.0678.793.7
ViT-VTNPre-train=IN-21K, GFLOPs=4218, Views=1 × 1, Params=11.02022.0678.693.7
DSA+I3DBackbone=3D ResNet-50, Sampling (Tinfer x Uinfer x #crop)=8x8x3, GFLOPs=167.7x2x3=10062021.0578.293.2
SlowfastBackbone=3D R101+3D R101, Sampling (Tinfer x Uinfer x #crop)=(8+32)x3x10, GFLOPs=106x30=31802021.0577.993.2
ip-CSN-152Pre-train=None, GFLOPs=109, Views=10×3, Params=32.82022.0677.892.8
NL I3D-101Pre-train=IN-1K, GFLOPs=359, Views=10×3, Params=61.82022.0677.793.3
DSA+I3DBackbone=3D ResNet-50, Sampling (Tinfer x Uinfer x #crop)=4x8x3, GFLOPs=83.8x2x3=5032021.0577.793.1
X3D-LBackbone=not specified, Sampling (Tinfer x Uinfer x #crop)=16x10x3, GFLOPs=24.8x30=7442021.0577.592.9
V4D+I3DBackbone=3D ResNet-50, Sampling (Tinfer x Uinfer x #crop)=8x10x3, GFLOPs=21462021.0577.493.1
MLP-3D-XSPre-train=IN-1K, GFLOPs=57, Views=4 × 3, Params=50.12022.0677.293.1
SlowfastBackbone=3D R50+3D R50, Sampling (Tinfer x Uinfer x #crop)=(8+32)x3x10, GFLOPs=65.7x30=19712021.057792.6
NL+I3DBackbone=3D ResNet-50, Sampling (Tinfer x Uinfer x #crop)=128x10x3, GFLOPs=282x30=84602021.0576.592.6
TANetBackbone=ResNet-50, Sampling (Tinfer x Uinfer x #crop)=8x10x3, GFLOPs=43x30=12902021.0576.192.3
MVFNetBackbone=ResNet-50, Sampling (Tinfer x Uinfer x #crop)=8x10x3, GFLOPs=33x30=9902021.057692.4
SlowfastBackbone=3D R50+3D R50, Sampling (Tinfer x Uinfer x #crop)=(4+32)x10x3, GFLOPs=36.1x30=10832021.0575.692.1
TEABackbone=ResNet-50, Sampling (Tinfer x Uinfer x #crop)=8x10x3, GFLOPs=35x30=10502021.057591.8
TEINetBackbone=ResNet-50, Sampling (Tinfer x Uinfer x #crop)=8x10x3, GFLOPs=33x30=9902021.0574.991.8
NL+I3DBackbone=3D ResNet-50, Sampling (Tinfer x Uinfer x #crop)=32x10x3, GFLOPs=70.5x30=21152021.0574.991.6
SlowonlyBackbone=3D ResNet-50, Sampling (Tinfer x Uinfer x #crop)=8x10x3, GFLOPs=41.9x30=12572021.0574.991.5
TSMBackbone=ResNet-50, Sampling (Tinfer x Uinfer x #crop)=8x10x3, GFLOPs=33x30=9902021.0574.191.2
I3DPre-train=IN-1K, GFLOPs=108, Params=25.02022.0672.190.3
R(2+1)DPre-train=None, GFLOPs=75, Views=10×1, Params=61.82022.067290