Real-time 3D Human Pose Estimation Through Augmented Reality Technology and Transpose Model

With the rapid development of virtual reality and augmented reality technologies, real-time 3D human pose estimation plays a key role in various interactive applications. This paper proposes a real-time 3D human pose estimation method based on augmented reality technology and the Transpose model. This method significantly improves the pose estimation accuracy in complex scenes by combining the scene perception module and the spatiotemporal feature module. We conducted extensive experimental verification on the Human3.6M dataset, and the results showed that the proposed method performed well in different activities and multiple perspectives. Through ablation experimental analysis, we further confirmed the contribution of each module to the overall performance. The final experimental results show that the complete model achieved the best performance in terms of MPJPE, MPVE and Angular error, verifying the effectiveness and stability of our method. This study provides an important reference for real-time human pose estimation in augmented reality and demonstrates its great potential in practical applications.