Research Article
Lite-3DCNN Combined with Attention Mechanism for Complex Human Movement Recognition
Table 3
The trainable parameters (in millions) of the proposed method and other methods for the UCF-101 dataset.
| Model type | 10 class (M) | 20 class (M) | 30 class (M) |
| C3D | 52.87 | 61.30 | 61.34 | Lite-3DCNN | 1.609 | 1.616 | 1.621 | Lite-3DCNN–LSTM (512) | 3.120 | 3.122 | 3.135 | Lite-3DCNN–BiLSTM (512) | 5.219 | 5.224 | 5.229 | Proposed method | 1.712 | 1.716 | 1.884 |
|
|