List your company
ResearchFormerly WizWand

Action Recognition on UAV Human (test)

55Top-1 AccuracyPMI Sampler
19.702428.866238.0347.1938Mar 21, 2022May 24, 2022Jul 28, 2022Oct 1, 2022Dec 5, 2022Feb 8, 2023Apr 14, 2023
Updated 8mo ago

Evaluation results

MethodLinks
PMI SamplerFrames Number=16, Input Size=224 x 224, Init.=Kinetics2023.0455
MG SamplerFrames Number=16, Input Size=224 x 224, Init.=Kinetics2023.0453.8
PMI SamplerFrames Number=8, Input Size=620 x 620, Init.=Kinetics2023.0452
MITFASBackbone=X3D-M, Frames Number=16, Input Size=224 x 224, Initialization=Kinetics2023.0350.8
MITFASFrames Number=16, Input Size=224 x 224, Init.=Kinetics2023.0450.8
PMI SamplerFrames Number=8, Input Size=540 x 540, Init.=Kinetics2023.0447.7
AZTRFrames Number=16, Input Size=224 x 224, Init.=Kinetics2023.0447.4
MITFASBackbone=X3D-M, Frames Number=8, Input Size=620 x 620, Initialization=Kinetics2023.0346.6
MITFASFrames Number=8, Input Size=620 x 620, Init.=Kinetics2023.0446.6
DiffFARFrames Number=8, Input Size=540 x 540, Init.=Kinetics2023.0441.9
MITFASBackbone=X3D-M, Frames Number=16, Input Size=224 x 224, Initialization=None2023.0340.2
PMI SamplerFrames Number=8, Input Size=540 x 540, Init.=None2023.0439.7
FARBackbone=X3D-M, Frames Number=8, Input Size=620 x 620, Initialization=Kinetics2023.0339.1
FARFrames Number=8, Input Size=620 x 620, Init.=Kinetics2023.0439.1
X3D-M + FARParam (M)=3.8, FLOPs (GFlops/video)=14.41, GPU (GB/video)=3, Inference Time (sec/video)=0.092022.0338.6
FARFrames Number=8, Input Size=540 x 540, Init.=Kinetics2023.0438.6
MITFASBackbone=X3D-M, Frames Number=8, Input Size=540 x 540, Initialization=None2023.0338.4
MITFASFrames Number=8, Input Size=540 x 540, Init.=None2023.0438.4
X3D-MParam (M)=3.8, FLOPs (GFlops/video)=14.39, GPU (GB/video)=3, Inference Time (sec/video)=0.082022.0336.6
X3D-MFrames Number=8, Input Size=540 x 540, Init.=Kinetics2023.0436.6
ViT-B-TimeSformerParam (M)=121.4, FLOPs (GFlops/video)=2380, GPU (GB/video)=32, Inference Time (sec/video)=0.272022.0333.9
FARBackbone=X3D-M, Frames Number=16, Input Size=224 x 224, Initialization=Kinetics2023.0331.9
FARFrames Number=16, Input Size=224 x 224, Init.=Kinetics2023.0431.9
X3D-MBackbone=-, Frames Number=16, Input Size=224 x 224, Initialization=Kinetics2023.0330.6
X3D-MFrames Number=16, Input Size=224 x 224, Init.=Kinetics2023.0430.6
I3D + FARParam (M)=12, FLOPs (GFlops/video)=346.6, GPU (GB/video)=10, Inference Time (sec/video)=0.22022.0329.21
FARBackbone=I3D, Frames Number=8, Input Size=540 x 960, Initialization=Kinetics2023.0329.2
FARBackbone=X3D-M, Frames Number=8, Input Size=540 x 540, Initialization=None2023.0328.8
FARFrames Number=8, Input Size=540 x 540, Init.=None2023.0428.8
X3D-LBackbone=-, Frames Number=16, Input Size=224 x 224, Initialization=None2023.0327.6
FARBackbone=X3D-M, Frames Number=16, Input Size=224 x 224, Initialization=None2023.0327.6
X3D-MBackbone=-, Frames Number=16, Input Size=224 x 224, Initialization=None2023.0327
I3D + FNetParam (M)=12, FLOPs (GFlops/video)=346.56, GPU (GB/video)=10, Inference Time (sec/video)=0.12022.0324.39
MVITParam (M)=36.6, FLOPs (GFlops/video)=70.8, GPU (GB/video)=9, Inference Time (sec/video)=0.162022.0324.3
FNetBackbone=I3D, Frames Number=8, Input Size=540 x 960, Initialization=Kinetics2023.0324.3
MVITBackbone=-, Frames Number=16, Input Size=224 x 224, Initialization=Kinetics2023.0324.3
I3D + Fourier AttentionParam (M)=12, FLOPs (GFlops/video)=346.57, GPU (GB/video)=10, Inference Time (sec/video)=0.192022.0324.15
I3D + Efficient AttentionParam (M)=12, FLOPs (GFlops/video)=462, GPU (GB/video)=13.3, Inference Time (sec/video)=0.122022.0321.13
I3DBackbone=ResNet-101, Frames Number=8, Input Size=540 x 960, Initialization=Kinetics2023.0321.1
I3D-MParam (M)=12, FLOPs (GFlops/video)=346.55, GPU (GB/video)=10, Inference Time (sec/video)=0.12022.0321.06