Table 4 Ablation study of dual-stream approach on X-View of NTU RGB + D 60.
From: Two-stream spatio-temporal GCN-transformer networks for skeleton-based action recognition
Model | Bones | Accuracy (%) | FLOPs (×109) | # Param. (×106) |
|---|---|---|---|---|
Gnet | Â | 94.35 | 21.59 | 3.97 |
Tnet | Â | 52.49 | 14.86 | 2.85 |
SA-TDGFormer | Â | 95.04 | 21.59 | 3.97 |
Gnet | √ | 96.36 | 32.78 | 7.36 |
Tnet | √ | 53.67 | 25.24 | 4.33 |
SA-TDGFormer | √ | 96.83 | 32.78 | 7.36 |
