Description of Invention
Human Pose Estimation (HPE) in computer vision is crucial but challenging, often hindered by low-resolution outputs and insufficient contextual awareness in key-point detection. This work addresses these issues by evaluating the high-resolution network (HRNet) and introducing a dual self-attention (DSA) mechanism to enhance global awareness. Integrating DSA into HRNet creates DSA-HRNet, which improves HPE performance. Tested on the COCO Val 2017 dataset, DSA-HRNet shows a 2.3% increase in mean average precision, 3% in AP at 50, and 2.7% in AP at 75, offering a streamlined solution for better HPE.