In the traditional training and teaching process of badminton, coaches usually use the traditional way to teach, and this type of teaching method is prone to many problems such as high subjectivity and lack of data support. In this study, by incorporating the Attention Mechanism module to calculate the model loss with Euclidean distance, we propose a Combined Attention Mechanism Network with Euclidean Distance (CPAE-HRNet), which enhances the model’s ability to process and learn spatial and image details. Experiments on the COCO dataset and the self-constructed badminton action pose dataset show some improvement in the performance of CPAE-HRNet.