Skeleton-based Action Recognition on NTU 120 (X-sub)
92Accuracy3Mformer
Evaluation results
| Method | Links | ||
|---|---|---|---|
| 3MformerPooling strategy=rank-pool, Method Category=Transformer-based2023.03 | 92 | — | |
| 3MformerPooling strategy=tri-pool, Method Category=Transformer-based2023.03 | 91.2 | — | |
| TSGCNeXt2023.04 | 90.22 | — | |
| HD-GCN2023.04 | 90.1 | — | |
| MAMPBackbone=Transformer, Evaluation Protocol=fine-tuned2023.08 | 90 | — | |
| MAMPEvaluation Protocol=fully supervised, Multi-stream ensembling=false2023.08 | 90 | — | |
| LST2023.04 | 89.9 | — | |
| InfoGCNVenue=CVPR'22, Method Category=Graph-based2023.03 | 89.8 | — | |
| InfoGCN* w/SkeletonGCLModality=4S2023.01 | 89.8 | — | |
| InfoGCNEvaluation Protocol=fully supervised2023.08 | 89.8 | — | |
| 3MformerPooling strategy=attn-pool, Method Category=Transformer-based2023.03 | 89.7 | — | |
| CTR-GCN* w/SkeletonGCLModality=4S2023.01 | 89.5 | — | |
| PSUMNet2023.04 | 89.4 | — | |
| TCA-GCN2023.04 | 89.4 | — | |
| CTR-GCNModality=4S2023.01 | 89.4 | — | |
| InfoGCNModality=4S2023.01 | 89.4 | — | |
| CTR-GCN* w/SkeletonGCLModality=J+B2023.01 | 89.2 | — | |
| InfoGCN*Modality=4S2023.01 | 89.2 | — | |
| CTR-GCNSkeleton format=3D2022.05 | 88.9 | — | |
| CTR-GCNVenue=ICCV'21, Method Category=Graph-based2023.03 | 88.9 | — | |
| CTR-GCN*Modality=4S2023.01 | 88.9 | — | |
| CTR-CGNEvaluation Protocol=fully supervised2023.08 | 88.9 | — | |
| InfoGCN* w/SkeletonGCLModality=J+B2023.01 | 88.8 | — | |
| STGAT2023.04 | 88.7 | — | |
| 2S-AGCN* w/SkeletonGCLModality=4S2023.01 | 88.7 | — | |
| CTR-GCN*Modality=J+B2023.01 | 88.7 | — | |
| PYSKLSkeleton format=3D2022.05 | 88.6 | — | |
| CTR-GCNModality=J+B2023.01 | 88.5 | — | |
| InfoGCNModality=J+B2023.01 | 88.5 | — | |
| 2S-AGCN* w/SkeletonGCLModality=J+B2023.01 | 88.4 | — | |
| 2S-AGCN*Modality=4S2023.01 | 88.2 | — | |
| InfoGCN*Modality=J+B2023.01 | 88.2 | — | |
| 3MformerPooling strategy=avg-pool, Method Category=Transformer-based2023.03 | 88 | — | |
| FreqMixFormerArchitecture Type=Transformer, Param (M)=2.04, GFLOPs=2.402024.07 | 87.9 | — | |
| 2S-AGCN*Modality=J+B2023.01 | 87.8 | — | |
| MST-GCNModality=4S2023.01 | 87.5 | — | |
| InfoGCN* w/SkeletonGCLModality=B2023.01 | 87.4 | — | |
| CTR-GCNModality=B2023.01 | 87.3 | — | |
| InfoGCNModality=B2023.01 | 87.3 | — | |
| HyperformerArchitecture Type=Transformer, Param (M)=2.69, GFLOPs=9.642024.07 | 87.3 | — | |
| SkeMixFormerArchitecture Type=Transformer, Param (M)=2.08, GFLOPs=2.392024.07 | 87.1 | — | |
| SD-HGCNVenue=ICONIP'21, Method Category=Hypergraph-based2023.03 | 87 | — | |
| MST-GCNModality=J+B2023.01 | 87 | — | |
| SkeletonMAE*Backbone=Transformer, Evaluation Protocol=fine-tuned2023.08 | 87 | — | |
| MS-G3D Net2020.03 | 86.9 | — | |
| MS-G3DVenue=CVPR'20, Method Category=Graph-based2023.03 | 86.9 | — | |
| PoseConv3DVenue=CVPR'22, Method Category=Graph-based2023.03 | 86.9 | — | |
| MS-G3DModality=J2023.01 | 86.9 | — | |
| CTR-GCN* w/SkeletonGCLModality=B2023.01 | 86.9 | — | |
| InfoGCN*Modality=B2023.01 | 86.9 | — | |
| DSTA-Net2020.07 | 86.6 | — | |
| LA-GCNGFLOPs=1.76, Param (M)=1.462023.05 | 86.5 | — | |
| DC-GCN+ADGModality=J2023.01 | 86.5 | — | |
| 4s-GSTNVenue=Symmetry'22, Method Category=Transformer-based2023.03 | 86.4 | — | |
| MTTVenue=LSP'21, Method Category=Transformer-based2023.03 | 86.1 | — | |
| DHGCNVenue=CoRR'21, Method Category=Hypergraph-based2023.03 | 86 | — | |
| 2S-AGCN* w/SkeletonGCLModality=B2023.01 | 86 | — | |
| PoseC3DEvaluation Protocol=fully supervised2023.08 | 86 | — | |
| Shift-GCNVenue=CVPR'20, Method Category=Graph-based2023.03 | 85.9 | — | |
| Shift-GCNModality=4S2023.01 | 85.9 | — | |
| CTR-GCN*Modality=B2023.01 | 85.7 | — | |
| HD-GCNArchitecture Type=GCN, Param (M)=1.68, GFLOPs=1.602024.07 | 85.7 | — | |
| CTR-GCN* w/SkeletonGCLModality=J2023.01 | 85.6 | — | |
| FGCN2020.03 | 85.4 | — | |
| Hi-TRSBackbone=3s-Transformer, Evaluation Protocol=fine-tuned2023.08 | 85.3 | — | |
| InfoGCN* w/SkeletonGCLModality=J2023.01 | 85.2 | — | |
| ST-TRVenue=CVIU'21, Method Category=Transformer-based2023.03 | 85.1 | — | |
| InfoGCNGFLOPs=1.84, Param (M)=1.572023.05 | 85.1 | — | |
| 2S-AGCN*Modality=B2023.01 | 85.1 | — | |
| CTR-GCNModality=J2023.01 | 85.1 | — | |
| InfoGCNModality=J2023.01 | 85.1 | — | |
| InfoGCNArchitecture Type=GCN, Param (M)=1.57, GFLOPs=1.682024.07 | 85.1 | — | |
| ViABackbone=2s-UINK, Evaluation Protocol=fine-tuned2023.08 | 85 | — | |
| MS-G3DGFLOPs=5.22, Param (M)=3.222023.05 | 84.9 | — | |
| CTR-GCNGFLOPs=1.97, Param (M)=1.462023.05 | 84.9 | — | |
| CTR-GCN*Modality=J2023.01 | 84.9 | — | |
| MS-G3DArchitecture Type=GCN, Param (M)=3.22, GFLOPs=5.222024.07 | 84.9 | — | |
| CTR-GCNArchitecture Type=GCN, Param (M)=1.46, GFLOPs=1.972024.07 | 84.9 | — | |
| MST-GCNModality=B2023.01 | 84.8 | — | |
| ST-GCN++2023.04 | 84.7 | — | |
| 2S-AGCN* w/SkeletonGCLModality=J2023.01 | 84.7 | — | |
| InfoGCN*Modality=J2023.01 | 84.2 | — | |
| DC-GCNGFLOPs=1.83, Param (M)=3.372023.05 | 84 | — | |
| 2S-AGCN*Modality=J2023.01 | 84 | — | |
| Shift-GCNModality=J+B2023.01 | 83.2 | — | |
| Shap-MixImbalance Factor (IF)=10, Stream configuration=single joint stream2024.07 | 83 | — | |
| 2s-AGCN2020.03 | 82.9 | — | |
| MST-GCNModality=J2023.01 | 82.8 | — | |
| 2S-AGCNModality=J2023.01 | 82.8 | — | |
| ST-TR-GCNModality=J2023.01 | 82.7 | — | |
| 2S-AGCNVenue=CVPR'19, Method Category=Graph-based2023.03 | 82.5 | — | |
| GLMCImbalance Factor (IF)=10, Stream configuration=single joint stream2024.07 | 82 | — | |
| Remix-DRWImbalance Factor (IF)=10, Stream configuration=single joint stream2024.07 | 81.6 | — | |
| MCCBackbone=2s-AGCN, Evaluation Protocol=fine-tuned2023.08 | 81.3 | — | |
| AMRprotocol=Linear evaluation2026.06 | 81.1 | — | |
| Shift-GCNModality=J2023.01 | 80.9 | — | |
| Balanced SoftmaxImbalance Factor (IF)=10, Stream configuration=single joint stream2024.07 | 80.7 | — | |
| CrosSCLRBackbone=3s-ST-GCN, Evaluation Protocol=fine-tuned2023.08 | 80.5 | — | |
| AimCLRBackbone=3s-ST-GCN, Evaluation Protocol=fine-tuned2023.08 | 80.1 | — | |
| BaselineImbalance Factor (IF)=10, Stream configuration=single joint stream2024.07 | 79.6 | — |