Action Recognition on NTU RGB+D 120 (X-set)
97AccuracyPAN-Ensemble
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PAN-Ensemble2D Skeleton=true, 3D Skeleton=true, RGB=true2025.12 | 97 | — | — | |
| BHaRNet-MModality=J+B+JM+BM+RGB, GFLOPs=13.0, Params(M)=22.22026.01 | 96.5 | — | — | |
| PoseC3DPre-training=true, Feature: RGB=true, Feature: Est. Pose=true, Feature: Annot. Pose=false2022.10 | 96.4 | — | — | |
| PoseC3DModality=RGB + Estimated Pose, Pre-trained=true2022.12 | 96.4 | — | — | |
| PoseConv3DModality=B+RGB, GFLOPs=41.8*, Params(M)=31.6*2026.01 | 96.4 | — | — | |
| SkeleTGFLOPs=9.6, Params(M)=5.22026.01 | 96.4 | — | — | |
| PoseC3D2D Skeleton=re-estimated, 3D Skeleton=false, RGB=true2025.12 | 96.4 | — | — | |
| BHaRNet-MModality=J+B+RGB, GFLOPs=7.6, Params(M)=16.72026.01 | 96.2 | — | — | |
| π-ViT (late fusion)2D Skeleton=training-only, 3D Skeleton=true, RGB=true2025.12 | 96.1 | — | — | |
| BHaRNet-M†Modality=J+B+RGB, GFLOPs=7.6, Params(M)=16.72026.01 | 96 | — | — | |
| Ours (BHaRNet-P)GFLOPs=6.8, Params(M)=9.72026.01 | 95.8 | — | — | |
| Ours (BHaRNet-E)GFLOPs=10.9, Params(M)=112026.01 | 95.7 | — | — | |
| BHaRNet-E‡GFLOPs=3.3, Params(M)=2.82026.01 | 95.5 | — | — | |
| BHaRNet-P‡GFLOPs=3.4, Params(M)=4.92026.01 | 95.5 | — | — | |
| Ours (BHaRNet-B)GFLOPs=6.6, Params(M)=5.52026.01 | 95.5 | — | — | |
| SequentialPointNetReference=IEEE TII 222026.06 | 95.4 | — | — | |
| BHaRNet-P†GFLOPs=3.4, Params(M)=4.92026.01 | 95.2 | — | — | |
| BHaRNet-B‡GFLOPs=9.6, Params(M)=5.22026.01 | 95.2 | — | — | |
| HyperPointNetReference=ICME 222026.06 | 95.1 | — | — | |
| BHaRNet-E†GFLOPs=5.4, Params(M)=5.52026.01 | 95 | — | — | |
| BHaRNet-B†GFLOPs=3.3, Params(M)=2.82026.01 | 94.9 | — | — | |
| PST-Transformer MotionPDE2026.06 | 94.42 | — | — | |
| MMNetModality=RGB + Pose, Pre-trained=true2022.12 | 94.4 | — | — | |
| MMNetModality=J+B+RGB, GFLOPs=89.2, Params(M)=34.22026.01 | 94.4 | — | — | |
| EPAM-NetModality=J+RGB, GFLOPs=8.1, Params(M)=2.52026.01 | 94.3 | — | — | |
| PSTNet++ MotionPDE2026.06 | 94.21 | — | — | |
| PST-TransformerReference=IEEE TPAMI 232026.06 | 94 | — | — | |
| PAN (Guided)2D Skeleton=true, 3D Skeleton=false, RGB=true2025.12 | 93.9 | — | — | |
| PSTNet MotionPDE2026.06 | 93.9 | — | — | |
| PSTNetInput=point2022.05 | 93.8 | — | — | |
| 3MformerGFLOPs=58.5, Params(M)=6.72026.01 | 93.8 | — | — | |
| 3Mformer2D Skeleton=false, 3D Skeleton=true, RGB=false2025.12 | 93.8 | — | — | |
| PAN (Even)2D Skeleton=false, 3D Skeleton=false, RGB=true2025.12 | 93.8 | — | — | |
| PSTNetReference=ICLR 212026.06 | 93.8 | — | — | |
| PSTNet++Reference=IEEE TPAMI 222026.06 | 93.8 | — | — | |
| 3DV-PointNet++Input=voxel + point2022.05 | 93.5 | — | — | |
| 3DV-PointNet++Reference=CVPR 202026.06 | 93.5 | — | — | |
| P4TransformerReference=CVPR 212026.06 | 93.5 | — | — | |
| PAN-Unified2D Skeleton=true, 3D Skeleton=true, RGB=true2025.12 | 93 | — | — | |
| pi-ViTModality=J+RGB, GFLOPs=590.0, Params(M)=121.42026.01 | 92.9 | — | — | |
| π-ViT2D Skeleton=training-only, 3D Skeleton=training-only, RGB=true2025.12 | 92.9 | — | — | |
| EPP-NetModality=Skeleton+Parsing2024.01 | 92.8 | — | — | |
| REASON+LLMCategory=LLM, Mod.=J, Two-stage=N, Interpret.=Y2026.05 | 92.8 | — | — | |
| STAR-TransformerPre-training=false, Feature: RGB=true, Feature: Est. Pose=false, Feature: Annot. Pose=true2022.10 | 92.7 | — | — | |
| STARModality=RGB + Pose, Pre-trained=false2022.12 | 92.7 | — | — | |
| STAR-TransformerModality=Skeleton+RGB, Source=WACV'232024.01 | 92.7 | — | — | |
| EPP-Net (J+B+P)Modality=Skeleton+Parsing2024.01 | 92.7 | — | — | |
| STAR-TransformerFramework=Transformer2024.07 | 92.7 | — | — | |
| STAR-Transformer2D Skeleton=true, 3D Skeleton=false, RGB=true2025.12 | 92.7 | — | — | |
| X3D-UGTGFLOPs=9.15, Params (M)=0.96, Modality=RGB-only2026.02 | 92.67 | — | — | |
| 3DV (motion)Input=voxel2022.05 | 92.5 | — | — | |
| VPN++ (3D pose)Modality=RGB + Pose, Pre-trained=true2022.12 | 92.5 | — | — | |
| VPN++ (late fusion)2D Skeleton=false, 3D Skeleton=true, RGB=true2025.12 | 92.5 | — | — | |
| PartialVisGraphParams (M)=4.0, GFLOPs=2.90, Ensemble=4-stream2026.07 | 92.3 | — | — | |
| ProtoGCNEnsemble size=62024.11 | 92.2 | — | — | |
| ProtoGCNGFLOPs=43.4, Params(M)=24.92026.01 | 92.2 | — | — | |
| ProtoGCN2D Skeleton=false, 3D Skeleton=true, RGB=false2025.12 | 92.2 | — | — | |
| DeGCNGFLOPs=6.9, Params(M)=5.62026.01 | 92.1 | — | — | |
| Video Swin2D Skeleton=false, 3D Skeleton=false, RGB=true2025.12 | 92.1 | — | — | |
| Hyper-GCNCategory=GCN, Mod.=J+B+JM+BM, Two-stage=N, Interpret.=N2026.05 | 92 | — | — | |
| Hyper-GCNPublication=ICCV 2025, Params (M)=2.3, GFLOPs=2.88, Ensemble=4-stream2026.07 | 92 | — | — | |
| FreqMixFormerFramework=Transformer, Stream Configuration=6-stream2024.07 | 91.9 | — | — | |
| ProtoGCNEnsemble size=42024.11 | 91.9 | — | — | |
| HiODBackbone=GCN, Evaluation Protocol=FT.2026.02 | 91.9 | — | — | |
| REASONCategory=GCN, Mod.=J+B+JM+BM, Two-stage=N, Interpret.=Y2026.05 | 91.9 | — | — | |
| ProtoGCNPublication=CVPR 2025, Ensemble=4-stream2026.07 | 91.9 | — | — | |
| LA-GCNEnsemble=6-stream2023.05 | 91.8 | — | — | |
| LA-GCNModality=Skeleton+Text, Source=arXiv'232024.01 | 91.8 | — | — | |
| TSGCNeXt (new4s)Streams=4s2023.04 | 91.74 | — | — | |
| Shap-MixYear=-, Number of streams=42024.07 | 91.7 | — | — | |
| MMCLTraining Modality=Multi-modality, Inference Modality=Pose, Multi-stream ensemble=true2024.07 | 91.7 | — | — | |
| JT-GraphFormer2024.11 | 91.7 | — | — | |
| DVANetModality=RGB2023.12 | 91.6 | — | — | |
| Shap-MixYear=-, Number of streams=22024.07 | 91.6 | — | — | |
| HD-GCNFramework=GCN, Stream Configuration=6-stream2024.07 | 91.6 | — | — | |
| HD-GCN2024.11 | 91.6 | — | — | |
| DVANetGFLOPs=25.5, Params (M)=40.5, Modality=RGB-only2026.02 | 91.6 | — | — | |
| TimeSformer2D Skeleton=false, 3D Skeleton=false, RGB=true2025.12 | 91.6 | — | — | |
| LLM-ARModality=Joint2024.03 | 91.5 | — | — | |
| FreqMixFormerFramework=Transformer, Stream Configuration=4-stream2024.07 | 91.5 | — | — | |
| BlockGCN2024.11 | 91.5 | — | — | |
| BlockGCNPublisher=CVPR20242025.01 | 91.5 | — | — | |
| BlockGCNBackbone=GCN, Evaluation Protocol=FT.2026.02 | 91.5 | — | — | |
| BlockGCNGFLOPs=6.5, Params(M)=5.22026.01 | 91.5 | — | — | |
| BlockGCN2D Skeleton=false, 3D Skeleton=true, RGB=false2025.12 | 91.5 | — | — | |
| BlockGCNCategory=GCN, Mod.=J+B+JM+BM, Two-stage=N, Interpret.=N2026.05 | 91.5 | — | — | |
| LLM-ARCategory=LLM, Mod.=J, Two-stage=Y, Interpret.=Y2026.05 | 91.5 | — | — | |
| BlockGCNPublication=CVPR 2024, Params (M)=1.3, GFLOPs=1.63, Ensemble=4-stream2026.07 | 91.5 | — | — | |
| TSGCNeXt (2s)Streams=2s2023.04 | 91.46 | — | — | |
| DG-STGCNPose Source=3D Skeletons, Fusion=Ensemble (4 modalities)2022.10 | 91.4 | — | — | |
| OursModality=RGB + Pose, Pre-trained=false2022.12 | 91.4 | — | — | |
| 3D-Def-Transformer2D Skeleton=true, 3D Skeleton=false, RGB=true2025.12 | 91.4 | — | — | |
| SkateFormer (4 ensemble)Mod.=All, Param. (M)=2.03, FLOPs (G)=14.48, Architecture paradigm=ANN-Trans.2026.03 | 91.4 | — | — | |
| SkateFormerCategory=Trans., Mod.=J+B+JM+BM, Two-stage=N, Interpret.=N2026.05 | 91.4 | — | — | |
| SkateFormerPublication=ECCV 2024, Params (M)=2.0, GFLOPs=3.62, Ensemble=4-stream2026.07 | 91.4 | — | — | |
| DG-STGCNPose Source=2D pose estimations, Fusion=Ensemble (4 modalities)2022.10 | 91.3 | — | — | |
| HyperformerType=Transformer, Parameters(M)=2.6M, Modality=4 streams2022.11 | 91.3 | — | — | |
| LA-GCNEnsemble=4-stream2023.05 | 91.3 | — | — | |
| STC-NetYear=ICCV’23, Number of streams=42024.07 | 91.3 | — | — | |
| KoopmanFramework=GCN2024.07 | 91.3 | — | — |