A Lightweight Gesture Recognition Network

This paper presents a lightweight gesture recognition algorithm model based on the YOLOv5s framework to address the challenges of large parameter size and high computational complexity in gesture recognition. The ShuffleNetV2 is used as the backbone network to reduce the computational load and improve detection speed. Lightweight modules such as GSConv and VoVGSCSP are introduced to further compress the model size while maintaining accuracy. The BiFPN structure is employed to enhance the detection accuracy of the network and reduce computational costs. The Coordinate Attention mechanism is introduced to enhance the network's focus on key features. Experimental results demonstrate that the proposed algorithm achieves an average precision of 95.2% on the HD-HaGRID dataset. Compared to the original YOLOv5s model, this model reduces the parameter count by 70.6% and the model size by 69.2%. Therefore, this model is suitable for real-time gesture recognition classification and detection, with significant potential for practical applications.