Experimental settings and hyperparameters.

Real-time pose estimation is essential in various applications such as sports analysis, motion tracking, and healthcare, where understanding human movement in complex environments is critical. However, existing methods often struggle to balance accuracy and computational efficiency, particularly in crowded or dynamic scenes where occlusions and fast movements are common. To address these challenges, we propose a novel architecture that integrates the Detection Transformer (DETR) with a Graph Convolutional Transformer (GCT) and Gating Mechanisms. This model is designed to capture both spatial and temporal dependencies more effectively while optimizing feature selection, leading to improved pose estimation accuracy and efficiency. Our approach outperforms current state-of-the-art methods, particularly in challenging scenarios, as demonstrated through extensive experiments on the PoseTrack Dataset. Experimental results show that the proposed model achieves superior performance across key metrics such as mean Average Precision (mAP), PCK@0.5, and Recall, while maintaining real-time processing capabilities. This research contributes to the field by offering a more robust solution for pose estimation in real-world, complex environments, with potential applications in sports analysis, surveillance, and human-robot interaction.