List your company
ResearchFormerly WizWand

Group Activity Recognition on Volleyball dataset (Accuracy)

95.4AccuracyDual-AI
61.70470.45279.287.948Aug 9, 2021Sep 17, 2021Oct 27, 2021Dec 6, 2021Jan 15, 2022Feb 24, 2022Apr 5, 2022
Updated 8mo ago

Evaluation results

MethodLinks
Dual-AIBackbone=Inception-v3, Data Ratio=100%, Optical Flow=True2022.0495.4
GFormerBackbone=I3D, Data Ratio=100%, Optical Flow=True2022.0494.9
Actor TransformersFeature extraction method=Pose + Optical flow2021.0894.4
Dual-AIBackbone=Inception-v3, Data Ratio=100%, Optical Flow=False2022.0494.4
Dual-AIBackbone=Inception-v3, Data Ratio=50%, Optical Flow=True2022.0494.2
GFormerBackbone=Inception-v3, Data Ratio=100%, Optical Flow=False2022.0494.1
ERNBackbone=R50-FPN+I3D, Data Ratio=100%, Optical Flow=True2022.0494.1
POGARS with ballFeature extraction method=Pose + Ball tracklets2021.0893.9
TCE+STBiPBackbone=Inception-v3, Data Ratio=100%, Optical Flow=False2022.0493.3
POGARS ALLFeature extraction method=Pose2021.0893.2
Joint learning with social groupsFeature extraction method=RGB2021.0893.1
DINBackbone=ResNet-18, Data Ratio=100%, Optical Flow=False2022.0493.1
JLSGBackbone=I3D, Data Ratio=100%, Optical Flow=True2022.0493.1
Convolutional Relational MachineFeature extraction method=RGB + Optical flow2021.0893
CRMBackbone=I3D, Data Ratio=100%, Optical Flow=True2022.0493
AformerBackbone=I3D, Data Ratio=100%, Optical Flow=True2022.0493
Dual-AIBackbone=Inception-v3, Data Ratio=50%, Optical Flow=False2022.0492.7
Action Relation GraphFeature extraction method=RGB2021.0892.6
ARGBackbone=Inception-v3, Data Ratio=100%, Optical Flow=False2022.0492.5
Actor TransformersFeature extraction method=Pose2021.0892.3
Group Interaction Relational NetworkFeature extraction method=Pose + Ball tracklets2021.0892.2
Spatio-temporal attention based modelFeature extraction method=Pose + RGB2021.0891.7
Dual-AIBackbone=Inception-v3, Data Ratio=25%, Optical Flow=True2022.0491.6
AFormerBackbone=I3D, Data Ratio=100%, Optical Flow=False2022.0491.4
SSUBackbone=Inception-v3, Data Ratio=100%, Optical Flow=False2022.0490.6
Multi-stream CNNFeature extraction method=RGB + Optical flow2021.0890.5
Social SceneFeature extraction method=RGB2021.0889.9
Dual-AIBackbone=Inception-v3, Data Ratio=25%, Optical Flow=False2022.0489.7
HRNBackbone=VGG19, Data Ratio=100%, Optical Flow=False2022.0489.5
StageNetBackbone=VGG16, Data Ratio=100%, Optical Flow=False2022.0489.3
Group Interaction Relational NetworkFeature extraction method=Pose2021.0888.4
I3DFeature extraction method=RGB2021.0884.6
CERNFeature extraction method=RGB2021.0883.3
CERNBackbone=VGG16, Data Ratio=100%, Optical Flow=False2022.0483.3
Two-stage hierarchical modelFeature extraction method=RGB2021.0881.9
HDTMBackbone=AlexNet, Data Ratio=100%, Optical Flow=False2022.0481.9
B2Feature extraction method=RGB2021.0875.1
B1Feature extraction method=RGB2021.0873.3
SBGARBackbone=Inception-v3, Data Ratio=100%, Optical Flow=True2022.0466.9
B3Feature extraction method=RGB2021.0863