A Static Letter Sign Language Recognition Algorithm Based on Improved YOLOv5

This article presents an improved static sign language recognition algorithm based on the YOLOv5 network model. The algorithm utilizes SimAM attention module and ASFF adaptive feature fusion module to optimize the original model and enhance its capacity for both feature extraction and target recognition. Experimental outcomes demonstrate that the enhanced model achieves a remarkable mAP increase of 3.7 percentage points while maintaining the same level of inference speed, manifesting an impressive level of effectiveness.