Research Article
Study of Human Motion Recognition Algorithm Based on Multichannel 3D Convolutional Neural Network
Table 2
The relationship between optimizer and recognition rate in UCF-101 dataset.
| | Iterations | SGD | RMSPROP | Adam |
| | 10 | 0.316 | 0.198 | 0.379 | | 20 | 0.389 | 0.256 | 0.445 | | 30 | 0.412 | 0.367 | 0.518 | | 40 | 0.532 | 0.467 | 0.593 | | 50 | 0.564 | 0.528 | 0.639 | | 60 | 0.589 | 0.698 | 0.717 | | 70 | 0.614 | 0.756 | 0.778 | | 80 | 0.698 | 0.779 | 0.801 | | 90 | 0.763 | 0.835 | 0.863 | | 100 | 0.838 | 0.867 | 0.916 |
|
|