Table 5.

Comparison of how different attention mechanisms performed on the noisy dataset (80% clear and 20% blurred images) (image tested on GTX 1660 SUPER) classification accuracy of similar gestures

Network structure	Precision	Recall	mAP 0.5	mAP 0.5:0.95
YOLOV5-P6	0.915	0.902	0.887	0.761
YOLOV5-P6+Transformer	0.921	0.9	0.891	0.759
YOLOV5-P6 +SE layer	0.927	0.914	0.903	0.765