摘要
本研究旨在通过引入注意力机制和优化损失函数,实现基于PoseNet模型的手势识别性能的提升。我们选择了MPII Human Pose数据集作为实验平台,该数据集提供了全身姿态估计的信息,通过数据处理将研究焦点集中在手部及其局部特征上,从而实现手势识别工作的评估。实验结果显示,在PCK和mAP等评价指标下,改进模型的性能得到了一定的提升;同时,模型在处理复杂环境条件下的稳定性和实时性也得到了增强通过数据分析和实证验证。
This study aims to improve the performance of gesture recognition based on the PoseNet model by incorporating attention mechanisms and optimizing the loss function.The MPII Human Pose dataset was chosen as the experimental platform,providing comprehensive information on full-body pose estimation.Through data processing,the research focus was directed towards the hand and its local features,thereby facilitating the evaluation of the gesture recognition task.Experimental results show that the performance of the improved model has been enhanced,as indicated by evaluation metrics such as PCK and mAP.Additionally,the model's stability and real-time capabilities under complex environmental conditions have also been strengthened.Through data analysis and empirical validation.
作者
赵佳娜
赵建光
Zhao Jiana;Zhao Jianguang(Information Engineering College,Hebei University of Architecture,Zhangjiakou,China)
出处
《科学技术创新》
2025年第18期105-108,共4页
Scientific and Technological Innovation
