Action Recognition on NTU-120 (Cross-Subject (XSub))
95.3AccuracyRGBPose-Conv3D
Evaluation Results
| Method | Links | |
|---|---|---|
| RGBPose-Conv3DModality=RGB + Pose, Fusion Strategy=early+late fusion, Backbone=ResNet-502021.04 | 95.3 | |
| PoseC3DModality=RGB + Estimated Pose, Pre-trained=true2022.12 | 95.3 | |
| E2E-GNetCategory=Deep Riemannian, Params (M)=0.93, FLOPs (G)=0.012026.03 | 95.2 | |
| MMNetModality=RGB + Pose, Pre-trained=true2022.12 | 92.9 | |
| AutoregAd-HGformerCategory=HyperGraph Transformer, Venue=WACV 2025, Params (M)=3.20, FLOPs (G)=15.42026.03 | 91 | |
| DeGCNParams (M)=5.62026.03 | 91 | |
| ProtoGCNCategory=Graph Convolution, Venue=CVPR 20252026.03 | 90.9 | |
| ProtoGCNSupervision Type=Supervised, Publication=CVPR 20252026.04 | 90.9 | |
| Method [12]Modality=RGB + Pose2021.04 | 90.7 | |
| VPN++ (3D pose)Modality=RGB + Pose, Pre-trained=true2022.12 | 90.7 | |
| KShapeNetCategory=Deep Riemannian, Venue=ICCV 2021, Params (M)=0.93, FLOPs (G)=0.012026.03 | 90.6 | |
| NDMHICSupervision Type=Supervised, Publication=AAAI 20252026.04 | 90.6 | |
| OursModality=RGB + Pose, Pre-trained=false2022.12 | 90.5 | |
| FreqMixFormerCategory=Transformer/ Graph Transformer, Venue=ACM MM 2024, Params (M)=2.04, FLOPs (G)=2.402026.03 | 90.5 | |
| AdaptiveHGCNCategory=HyperGraph Convolution, Venue=ICCV 2025, Params (M)=1.10, FLOPs (G)=1.632026.03 | 90.5 | |
| FreqMixFormerParams (M)=2.0, Flops (G)=64.42026.03 | 90.5 | |
| Shap-MixYear=-, Number of streams=42024.07 | 90.4 | |
| STARModality=RGB + Pose, Pre-trained=false2022.12 | 90.3 | |
| BlockGCNCategory=Graph Convolution, Venue=CVPR 2024, Params (M)=1.30, FLOPs (G)=1.632026.03 | 90.3 | |
| BlockGCNParams (M)=1.3, Flops (G)=1.62026.03 | 90.3 | |
| BlockGCNSupervision Type=Supervised, Publication=CVPR 2024, FLOPs/G=6.522026.04 | 90.3 | |
| VA-VRSupervision Type=Supervised, Publication=AAAI 20252026.04 | 90.3 | |
| Shap-MixYear=-, Number of streams=22024.07 | 90.2 | |
| 3s-S2IEnsemble=3-stream, Protocol=Fine-tuned2026.03 | 90.2 | |
| HD-GCNCategory=Graph Convolution, Venue=ICCV 2023, Params (M)=1.68, FLOPs (G)=1.602026.03 | 90.1 | |
| SkeMixFormerParams (M)=2.1, Flops (G)=4.82026.03 | 90.1 | |
| HD-GCNSupervision Type=Supervised, Publication=ICCV 2023, FLOPs/G=7.082026.04 | 90.1 | |
| MAMSCategory=Transformer/ Graph Transformer, Venue=ICCV 20232026.03 | 90 | |
| MAMPProtocol=Fine-tuned2026.03 | 90 | |
| + UKTLarchitecture=Tensor-based, backbone=MLP + HoT2026.03 | 90 | |
| STC-NetYear=ICCV’23, Number of streams=42024.07 | 89.9 | |
| GAPCategory=Graph Convolution, Venue=ICCV 20232026.03 | 89.9 | |
| JT-GraphFormerCategory=Transformer/ Graph Transformer, Venue=AAAI 20242026.03 | 89.9 | |
| HyperformerCategory=HyperGraph Transformer, Venue=ArXiv 2022, Params (M)=2.6, FLOPs (G)=14.82026.03 | 89.9 | |
| HyperformerParams (M)=2.7, Flops (G)=9.62026.03 | 89.9 | |
| DSDC-GCNarchitecture=Graph-based2026.03 | 89.9 | |
| GAPSupervision Type=Supervised, Publication=ICCV 20232026.04 | 89.9 | |
| JT-GraphFormerSupervision Type=Supervised, Publication=AAAI 20242026.04 | 89.9 | |
| InfoGCNModality=Pose, Pre-trained=false2022.12 | 89.8 | |
| Info-GCNYear=CVPR’22, Number of streams=62024.07 | 89.8 | |
| HD-GCNYear=ICCV’23, Number of streams=42024.07 | 89.8 | |
| InfoGCNCategory=Graph Convolution, Venue=CVPR 2022, Params (M)=1.60, FLOPs (G)=1.842026.03 | 89.8 | |
| SkeletonGCLCategory=Graph Convolution, Venue=ICLR 20232026.03 | 89.8 | |
| SkateFormerCategory=Transformer/ Graph Transformer, Venue=ECCV 2024, Params (M)=2.03, FLOPs (G)=3.622026.03 | 89.8 | |
| SkateFormerParams (M)=2.0, Flops (G)=3.62026.03 | 89.8 | |
| SkeletonGCLSupervision Type=Supervised, Publication=ICLR 20232026.04 | 89.8 | |
| StreamGCNYear=IJCAI’23, Number of streams=62024.07 | 89.7 | |
| CTR-GCNarchitecture=Graph-based2026.03 | 89.6 | |
| SBFConv3DCategory=Skl+SBF(J+L), Add. Anno.=✗, Params (M)=4.0, FLOPs (G)=32.4, Clips=102026.04 | 89.6 | |
| FR-HeadYear=CVPR’23, Number of streams=42024.07 | 89.5 | |
| FR HeadCategory=Graph Convolution, Venue=CVPR 2023, Params (M)=1.992026.03 | 89.5 | |
| FRHeadParams (M)=2.02026.03 | 89.5 | |
| FR-HeadSupervision Type=Supervised, Publication=CVPR 20232026.04 | 89.5 | |
| Info-GCNYear=CVPR’22, Number of streams=42024.07 | 89.4 | |
| TranSkeletonParams (M)=2.2, Flops (G)=9.22026.03 | 89.4 | |
| FD-GCNarchitecture=Graph-based2026.03 | 89.4 | |
| SBFConv3DCategory=Skl+SBF(J+L), Add. Anno.=✗, Params (M)=4.0, FLOPs (G)=32.4, Clips=12026.04 | 89.4 | |
| DS-GCNCategory=Graph Convolution, Venue=AAAI 20242026.03 | 89.2 | |
| InfoGCNParams (M)=1.6, Flops (G)=1.82026.03 | 89.2 | |
| DS-GCNSupervision Type=Supervised, Publication=AAAI 20242026.04 | 89.2 | |
| PSUMNetParams. (M)=3.22022.08 | 89.1 | |
| SBFConv3DCategory=Skl+SBF(J), Add. Anno.=✗, Params (M)=2.0, FLOPs (G)=16.2, Clips=102026.04 | 89.1 | |
| FG-STFormerCategory=Transformer/ Graph Transformer, Venue=ACCV 20222026.03 | 89 | |
| SBFConv3DCategory=Skl+SBF(J), Add. Anno.=✗, Params (M)=2.0, FLOPs (G)=16.2, Clips=12026.04 | 89 | |
| CTR-GCNYear=ICCV’21, Number of streams=42024.07 | 88.9 | |
| CTR-GCNCategory=Graph Convolution, Venue=ICCV 2021, Params (M)=1.46, FLOPs (G)=1.972026.03 | 88.9 | |
| CTR-GCNParams (M)=1.5, Flops (G)=2.02026.03 | 88.9 | |
| KGS-GCNParams (M)=1.4, Flops (G)=1.32026.03 | 88.9 | |
| DualHead-NetModality=Pose, Pre-trained=false2022.12 | 88.8 | |
| CTR-GCNModality=Pose, Pre-trained=false2022.12 | 88.8 | |
| DST-HCNCategory=HyperGraph Convolution, Venue=ICME 2023, Params (M)=2.93, FLOPs (G)=3.502026.03 | 88.8 | |
| + KTLarchitecture=Tensor-based, backbone=MLP + HoT2026.03 | 88.8 | |
| PYSKLCategory=Graph Convolution, Venue=ACM MM 2022, Params (M)=1.39, FLOPs (G)=2.802026.03 | 88.6 | |
| + KPCA (baseline)architecture=Tensor-based, backbone=MLP + HoT2026.03 | 88.6 | |
| CTR-GCNParams. (M)=5.62022.08 | 88.3 | |
| EfficientGCNYear=TPAMI’22, Number of streams=32024.07 | 88.3 | |
| EfficientGCNParams (M)=2.0, Flops (G)=15.22026.03 | 88.3 | |
| + TPCA (baseline)architecture=Tensor-based, backbone=MLP + HoT2026.03 | 88.2 | |
| MSICategory=Skl+Obj, Add. Anno.=✓2026.04 | 88.2 | |
| DSTA-NetParams. (M)=14.02022.08 | 87.8 | |
| Skeletal GNNModality=Pose, Pre-trained=false2022.12 | 87.5 | |
| MST-GCNYear=AAAI’21, Number of streams=42024.07 | 87.5 | |
| MST-GCNCategory=Graph Convolution, Venue=AAAI 2021, Params (M)=3.102026.03 | 87.5 | |
| ProtoGCNCategory=Skeleton, Add. Anno.=✗, Params (M)=24.9, FLOPs (G)=36.9, Clips=102026.04 | 87.5 | |
| MS-G3DCategory=Skeleton, Add. Anno.=✗, Params (M)=11.8, FLOPs (G)=27.4, Clips=102026.04 | 87.4 | |
| BlockGCNCategory=Skeleton, Add. Anno.=✗, Params (M)=5.2, FLOPs (G)=8.4, Clips=102026.04 | 87.3 | |
| MS-G3DParams. (M)=6.42022.08 | 87.1 | |
| TSMFModality=RGB + Pose, Pre-trained=false2022.12 | 87 | |
| SD-HGCNarchitecture=Hypergraph-based2026.03 | 87 | |
| Hyper-GCNarchitecture=Hypergraph-based2026.03 | 87 | |
| MS-G3DYear=CVPR’20, Number of streams=22024.07 | 86.9 | |
| MS-G3DCategory=Graph Convolution, Venue=CVPR 2020, Params (M)=3.17, FLOPs (G)=10.272026.03 | 86.9 | |
| MS-G3DParams (M)=2.8, Flops (G)=5.22026.03 | 86.9 | |
| MS-G3Darchitecture=Graph-based2026.03 | 86.9 | |
| PoseConv3DCategory=Skeleton, Add. Anno.=✗, Params (M)=4.0, FLOPs (G)=31.6, Clips=102026.04 | 86.9 | |
| S2IProtocol=Fine-tuned2026.03 | 86.6 | |
| CTR-GCNCategory=Skeleton, Add. Anno.=✗, Params (M)=5.7, FLOPs (G)=7.8, Clips=102026.04 | 86.6 | |
| DC-GCN+ADGCategory=HyperGraph Convolution, Venue=ECCV 2020, FLOPs (G)=16.202026.03 | 86.5 | |
| GeomNetCategory=Deep Riemannian, Venue=ICCV 20212026.03 | 86.5 | |
| PA-ResGCNParams. (M)=3.62022.08 | 86.4 |