Real-Time Multi-Scale Pose Semantic Recognition Algorithm Based on Lightweight Network

Aiming at the problems of high calculation complexity and insufficient real-time performance of ride attitude security recognition algorithm, an improved yolov8s algorithm is studied to realize lightweight network model and real-time multi-scale multiplier posture semantic recognition. Firstly, the lightweight multi-scale convolution structure is designed to improve the efficiency of the algorithm to adapt to the practical deployment; secondly, the adaptive parameter activation function is introduced into the head network to enhance the flexibility and adaptability of the network; and then designed for the key point coordinates and confidence degree output by the model network. Finally, the preliminary training of the model using the COCO data set shows that the improved mAP @0.5 is improved by 0.8%, and the model parameters are reduced by 3%; Randomly select 50 people for posture semantic experiment verification, and the results show that the algorithm meets the requirements of real-time monitoring of manned operation behavior safety.