Table 11 Accuracy comparison of ablation experiments.
From: Dynamic gesture recognition based on 2D convolutional neural network and feature fusion
Method | SE module | Top-1 accuracy | |
|---|---|---|---|
Northwestern | Cambridge | ||
Original frame + Optical flow frame | \(\checkmark\) | 97.64% | 98.62% |
Original frame + Optical flow frame | Â | 96.52% | 97.23% |
Original frame | \(\checkmark\) | 83.67% | 88.47% |
Original frame | Â | 81.44% | 87.18% |
Optical flow frame | \(\checkmark\) | 87.21% | 79.13% |
Optical flow frame | Â | 85.79% | 78.07% |
