Action Recognition on NTU RGB+D 60 (XSub & XView Protocols)
96.1Top-1 Acc (XSub)EPAM-Net (Our proposed approach)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| EPAM-Net (Our proposed approach)Modality=Multimodal (Skeleton+RGB), GFLOPs=8.1, Param (M)=2.52024.08 | 96.1 | 99 | 97.6 | |
| MMNet (Inception-v3)Modality=Multimodal (Skeleton+RGB), Backbone=Inception-v3, GFLOPs=89.2, Param (M)=34.22024.08 | 95.3 | 98.4 | 96.8 | |
| VPN++ + 3D PosesModality=Multimodal (Skeleton+RGB), GFLOPs=125.8, Param (M)=15.52024.08 | 94.9 | 98.1 | 96.5 | |
| TCEM-MMNetModality=Multimodal (Skeleton+RGB)2024.08 | 94.3 | 98.8 | 96.6 | |
| π-ViTModality=RGB, Skeleton usage=Training only, Input clips per video=10, GFLOPs=590.0, Param (M)=121.42024.08 | 94 | 97.9 | 96 | |
| PoseConv3DModality=Skeleton, GFLOPs=15.9, Param (M)=2.02024.08 | 93.7 | 96.6 | 95.2 | |
| VPN (I3D)Modality=Multimodal (Skeleton+RGB), Backbone=I3D, GFLOPs=107.9, Param (M)=24.02024.08 | 93.5 | 96.2 | 94.6 | |
| I3D-Resnet50Modality=RGB, Backbone=Resnet50, GFLOPs=51.7, Param (M)=33.02024.08 | 93.2 | 97.7 | 95.3 | |
| TSMFModality=Multimodal (Skeleton+RGB), GFLOPs=85.4, Param (M)=20.82024.08 | 92.5 | 97.4 | 95 | |
| STAR-TransformerModality=Multimodal (Skeleton+RGB)2024.08 | 92 | 96.5 | 94.3 | |
| VPN++Modality=RGB, Skeleton usage=Training only2024.08 | 91.9 | 94.9 | 93.4 | |
| MS-G3DModality=Skeleton, GFLOPs=16.7, Param (M)=2.82024.08 | 91.5 | 96.2 | 93.9 | |
| 2s-AGCNModality=Skeleton, GFLOPs=8.8, Param (M)=3.52024.08 | 88.5 | 95.1 | 91.8 | |
| ST-GCNModality=Skeleton, GFLOPs=3.8, Param (M)=3.12024.08 | 81.5 | 88.3 | 84.6 | |
| C3DModality=RGB, GFLOPs=38.5, Param (M)=78.42024.08 | 63.5 | 70.3 | 66.9 |