As the basic technology of human action recognition, pose estimation is attracting more and more researchers' attention, while edge application scenarios pose a higher challenge. The authors propose a lightweight multi-person pose estimation scheme to meet the needs of real-time human action recognition on the edge end. This scheme uses AlphaPose to extract human skeleton nodes and adds ResNet and Dense Upsampling Revolution to improve its accuracy. Meanwhile, YOLO is used to enhance AlphaPose’s support for multi-person pose estimation and to optimize the proposed model with TensorRT. In addition, the authors set Jetson Nano as the Edge AI deployment device of the proposed model and successfully realize the model migration to the edge end. The experimental results show that the speed of the optimized object detection model can reach 20 FPS, and the optimized multi-person pose estimation model can reach 10 FPS. With the image resolution of 320×240, the model’s accuracy is 73.2%, which can meet the real-time requirements. In short, our scheme can provide a basis for a lightweight multi-person action recognition scheme on the edge end.

• APA 7th style

• Chicago style

• IEEE style

• Vancouver style