Research on gesture recognition algorithm based on target detection

Gesture recognition technology is widely used in many fields, such as virtual reality, smart homes, auxiliary medical treatment, and human-computer interaction equipment. Through the analysis and understanding of human gestures, more natural and intuitive communication between humans and computers can be achieved. However, due to the complexity of hand motion in human-computer interaction, gesture recognition faces many challenges, such as self-occlusion, self-collision, and appearance similarity of the hand. In order to solve these problems, this study proposes a gesture recognition algorithm based on target detection. Dynamic Snake Convolution(DSC) and Adaptive Feature Pyramid Network(AFPN) modules are introduced into the You Only Look Once Version 8(YOLOv8) model. The experimental results show that the enhanced YOLOv8 model equipped with DSC and AFPN module performs better than baseline YOLOv8 in gesture recognition tasks. In addition, experiments and ablation studies on the Light-HaGRID dataset further prove the effectiveness of the improved YOLOv8 model and the contributions of its components.