Research on Gesture Recognition Optimization Based on YOLOv8 and Efficient Vision Transformer Network
Yaowu Xu, Daisheng Xu, Peng Yang, Haolong Wu · 2024
Gesture recognition technology is widely used in the fields of human-computer interaction, virtual reality, and smart home, but it still faces challenges of recognition accuracy, real-time performance, and adaptability to complex environments. Given the problems such as low precision rate and slow speed in practical applications, this study proposes an improved YOLOvS gesture recognition algorithm. Leap Motion was used to collect gesture data, and labeling was used to complete the gesture data set. The algorithm added an EfficientViT visual transformation network based on YOLOvS to improve the accuracy and speed of gesture recognition. The introduction of the Triplet Attention mechanism and Diou loss function enhances the feature extraction capability of gesture recognition and improves the accuracy and robustness of target detection. The results show that compared with the original YOLOvS algorithm, the proposed algorithm improves the accuracy by 4.5% and the detection speed by S.7%• The algorithm has strong practicability and can be applied to human-computer interaction and other fields.