Action Recognition on NTU-120 (Cross-Subject 1 (CS1))
90.7Top-1 AccuracyVPN++
Evaluation Results
| Method | Links | |
|---|---|---|
| VPN++Stream=VPN++ + 3D Poses2021.05 | 90.7 | |
| VPN++ + 3D PosesPose=true, RGB=true, Attention mechanism=true2021.05 | 90.7 | |
| VPN++ + 3D PosesStream=VPN++ + 3D Poses2021.05 | 90.7 | |
| MS-G3D NetPose=true, RGB=false, Attention mechanism=false2021.05 | 86.9 | |
| VPN++Stream=Distilled2021.05 | 86.7 | |
| VPN++Pose=false, RGB=true, Attention mechanism=true2021.05 | 86.7 | |
| VPN++Stream=VPN++2021.05 | 86.7 | |
| VPNStream=Attention Fusion2021.05 | 86.3 | |
| VPNPose=true, RGB=true, Attention mechanism=true2021.05 | 86.3 | |
| VPNStream=I1 + I2 (attention)2021.05 | 86.3 | |
| VPN-AStream=Distilled2021.05 | 85.2 | |
| VPN-AStream=VPN-A2021.05 | 85.2 | |
| VPN-FStream=Distilled2021.05 | 85.1 | |
| VPN-FStream=VPN-F2021.05 | 85.1 | |
| Separable STA*Pose=true, RGB=true, Attention mechanism=true2021.05 | 83.8 | |
| 2s-AGCNPose=true, RGB=false, Attention mechanism=true2021.05 | 82.9 | |
| I3D + AGCN-JStream=Late Fusion2021.05 | 81.1 | |
| RGB + 3D Poses (Late Fusion)Stream=I1 + I2 (Late Fusion)2021.05 | 81.1 | |
| AGCN-JStream=3D Poses2021.05 | 79.6 | |
| AGCN-JStream=3D Poses2021.05 | 79.6 | |
| I3DStream=RGB2021.05 | 77 | |
| I3D*Pose=false, RGB=true, Attention mechanism=false2021.05 | 77 | |
| I3DStream=RGB2021.05 | 77 | |
| PEMPose=false, RGB=true, Attention mechanism=true2021.05 | 64.6 | |
| Multi-Task CNNPose=false, RGB=false, Attention mechanism=true2021.05 | 62.2 | |
| 2s-Att LSTMPose=false, RGB=true, Attention mechanism=true2021.05 | 61.2 | |
| Two-streams + ST-LSTMPose=true, RGB=true, Attention mechanism=false2021.05 | 61.2 | |
| Two-streamsPose=false, RGB=true, Attention mechanism=false2021.05 | 58.5 |