Skeleton-based Action Recognition on Kinetics-Skeleton
53.6Top-1 AccSkeletonAgent
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SkeletonAgentEnsemble size=62025.11 | 53.6 | 77 | |
| SkeletonAgentEnsemble size=42025.11 | 53 | 76.8 | |
| SkeletonAgentEnsemble size=22025.11 | 52.2 | 75.9 | |
| ACLNet2026.01 | 52.1 | 75.9 | |
| ProtoGCNPublication=CVPR 20252025.11 | 51.9 | 75.6 | |
| DS-GCNPublication=AAAI 20242025.11 | 50.6 | — | |
| DS-GCN2026.01 | 50.6 | — | |
| PYSKLPublication=ACM MM 20222025.11 | 49.1 | — | |
| 3MformerPooling strategy=rank-pool, Method Category=Transformer-based2023.03 | 48.3 | 72.3 | |
| STFD-Net2026.01 | 48.1 | — | |
| PoseConv3DVenue=CVPR'22, Method Category=Graph-based2023.03 | 47.7 | — | |
| 3MformerPooling strategy=tri-pool, Method Category=Transformer-based2023.03 | 47.7 | 71.9 | |
| 3MformerPooling strategy=attn-pool, Method Category=Transformer-based2023.03 | 45.7 | 67.6 | |
| 3MformerPooling strategy=avg-pool, Method Category=Transformer-based2023.03 | 43.1 | 65.2 | |
| Lang2Motionevaluation_mode=zero-shot2025.12 | 41.6 | 64.3 | |
| HD-GCNPublication=ICCV 20232025.11 | 40.9 | 63.5 | |
| UPSPublication=CVPR 20232025.11 | 40.5 | 63.3 | |
| UPS2026.01 | 40.5 | 63.3 | |
| STFPublication=AAAI 20222025.11 | 39.9 | — | |
| + UKTLarchitecture=Tensor-based, backbone=MLP + HoT2026.03 | 39.2 | 62.3 | |
| + KTLarchitecture=Tensor-based, backbone=MLP + HoT2026.03 | 38.9 | 61.9 | |
| DSDC-GCNarchitecture=Graph-based2026.03 | 38.6 | 63.4 | |
| STSTVenue=ACM MM'21, Method Category=Transformer-based2023.03 | 38.3 | 61.2 | |
| STSTarchitecture=Transformer-based2026.03 | 38.3 | 61.2 | |
| MST-GCNPublication=AAAI 20212025.11 | 38.1 | 60.8 | |
| MS-G3DVenue=CVPR'20, Method Category=Graph-based2023.03 | 38 | 60.9 | |
| Selective-HCNVenue=ICMR'22, Method Category=Hypergraph-based2023.03 | 38 | 61.1 | |
| ST-TRVenue=CVIU'21, Method Category=Transformer-based2023.03 | 38 | 60.5 | |
| MS-G3DPublication=CVPR 20202025.11 | 38 | 60.9 | |
| MS-G3D2026.01 | 38 | 60.9 | |
| MS-G3D2025.12 | 38 | 60.9 | |
| MS-G3Darchitecture=Graph-based2026.03 | 38 | 60.9 | |
| Selective-HCNarchitecture=Hypergraph-based2026.03 | 38 | 61.1 | |
| ST-TRarchitecture=Transformer-based2026.03 | 38 | 60.5 | |
| + TPCA (baseline)architecture=Tensor-based, backbone=MLP + HoT2026.03 | 38 | 60.5 | |
| MTTVenue=LSP'21, Method Category=Transformer-based2023.03 | 37.9 | 61.3 | |
| MTTarchitecture=Transformer-based2026.03 | 37.9 | 61.3 | |
| MS-AAGCNNets=GCN2020.03 | 37.8 | 61 | |
| MS-AAGCNYear=2019, Architecture=GCN2020.03 | 37.8 | 61 | |
| DHGCNVenue=CoRR'21, Method Category=Hypergraph-based2023.03 | 37.7 | 60.6 | |
| DHGCNarchitecture=Hypergraph-based2026.03 | 37.7 | 60.6 | |
| UNFGEF2020.03 | 37.6 | 60.5 | |
| JB-AAGCNNets=GCN2020.03 | 37.4 | 60.4 | |
| JB-AAGCNYear=2019, Architecture=GCN2020.03 | 37.4 | 60.4 | |
| SD-HGCNVenue=ICONIP'21, Method Category=Hypergraph-based2023.03 | 37.4 | 60.5 | |
| SD-HGCNarchitecture=Hypergraph-based2026.03 | 37.4 | 60.5 | |
| Sym-GNNVenue=TPAMI'22, Method Category=Graph-based2023.03 | 37.2 | 58.1 | |
| Sym-GNNarchitecture=Graph-based2026.03 | 37.2 | 58.1 | |
| NAS-GCNInput Modality=Joint+Bone2019.11 | 37.1 | 60.1 | |
| GCN-NAS(Joint&Bone)Nets=GCN2020.03 | 37.1 | 60.1 | |
| GCN-NAS(Joint&Bone)Year=2019, Architecture=GCN2020.03 | 37.1 | 60.1 | |
| NAS-GCNVenue=AAAI'20, Method Category=Graph-based2023.03 | 37.1 | 60.1 | |
| Hyper-GNNVenue=TIP'21, Method Category=Hypergraph-based2023.03 | 37.1 | 60 | |
| NAS-GCNarchitecture=Graph-based2026.03 | 37.1 | 60.1 | |
| Hyper-GNNarchitecture=Hypergraph-based2026.03 | 37.1 | 60 | |
| + KPCA (baseline)architecture=Tensor-based, backbone=MLP + HoT2026.03 | 37.1 | 59.8 | |
| DGNNNets=GCN2020.03 | 36.9 | 59.6 | |
| DGNNYear=2019, Architecture=GCN2020.03 | 36.9 | 59.6 | |
| 2s-AGCNYear=2019, Architecture=GCN, Implementation=Trained by authors2020.03 | 36.8 | 59.2 | |
| Backbonearchitecture=Tensor-based, backbone=MLP + HoT2026.03 | 36.7 | 59.5 | |
| 2S-AGCNInput Modality=Joint+Bone, Conference=CVPR20192019.11 | 36.1 | 58.7 | |
| 2s-AGCN2020.03 | 36.1 | 58.7 | |
| 2S-AGCNVenue=CVPR'19, Method Category=Graph-based2023.03 | 36.1 | 58.7 | |
| 2s-AGCNPublication=CVPR 20192025.11 | 36.1 | 58.7 | |
| 2s-AGCN2026.01 | 36.1 | 58.7 | |
| 2S-AGCNarchitecture=Graph-based2026.03 | 36.1 | 58.7 | |
| NAS-GCNInput Modality=Joint2019.11 | 35.5 | 57.9 | |
| MotionCLIP2025.12 | 35.2 | 58.4 | |
| 2S-AGCNInput Modality=Joint, Conference=CVPR20192019.11 | 35.1 | 57.1 | |
| NAS-GCNInput Modality=Bone2019.11 | 34.9 | 57.1 | |
| AS-GCNConference=CVPR20192019.11 | 34.8 | 56.5 | |
| AS-GCNNets=GCN2020.03 | 34.8 | 56.5 | |
| PeGCN_totalNets=GCN2020.03 | 34.8 | 57.2 | |
| AS-GCNYear=2019, Architecture=GCN2020.03 | 34.8 | 56.5 | |
| PeGCNYear=2020, Architecture=GCN2020.03 | 34.8 | 57.24 | |
| AS-GCN2020.03 | 34.8 | 56.3 | |
| AS-GCNVenue=CVPR'19, Method Category=Graph-based2023.03 | 34.8 | 56.5 | |
| AS-GCNarchitecture=Graph-based2026.03 | 34.8 | 56.5 | |
| Js-AGCNNets=GCN, Mode=Backbone, Evaluation Protocol=self-evaluated2020.03 | 34.4 | 57 | |
| Js-AGCNYear=2019, Architecture=GCN, Implementation=Trained by authors2020.03 | 34.4 | 57.1 | |
| Bs-AGCNYear=2019, Architecture=GCN, Implementation=Trained by authors2020.03 | 34.1 | 57 | |
| 2S-AGCNInput Modality=Bone, Conference=CVPR20192019.11 | 33.3 | 55.7 | |
| ST-GCNNets=GCN, Evaluation Protocol=self-evaluated2020.03 | 31.6 | 53.7 | |
| ST-GCNYear=2018, Architecture=GCN, Implementation=Trained by authors2020.03 | 31.6 | 53.7 | |
| ST-GCNConference=AAAI20182019.11 | 30.7 | 52.8 | |
| ST-GCN2020.03 | 30.7 | 52.8 | |
| ST-GCNVenue=AAAI'18, Method Category=Graph-based2023.03 | 30.7 | 52.8 | |
| ST-GCNPublication=AAAI 20182025.11 | 30.7 | 52.8 | |
| ST-GCN2026.01 | 30.7 | 52.8 | |
| ST-GCN2025.12 | 30.7 | 52.8 | |
| ST-GCNarchitecture=Graph-based2026.03 | 30.7 | 52.8 | |
| TCNConference=CVPRW20172019.11 | 20.3 | 40 | |
| TCNYear=2017, Architecture=CNN2020.03 | 20.3 | 40 | |
| TCN2020.03 | 20.3 | 40 | |
| TCNVenue=CVPRW'17, Method Category=Graph-based2023.03 | 20.3 | 40 | |
| TCNarchitecture=Graph-based2026.03 | 20.3 | 40 | |
| P-LSTMConference=CVPR20162019.11 | 16.4 | 35.3 | |
| Deep LSTMYear=2016, Architecture=LSTM2020.03 | 16.4 | 35.3 | |
| Deep LSTM2020.03 | 16.4 | 35.3 | |
| FeatureConference=CVPR20152019.11 | 14.9 | 25.8 |