Research on Deep Learning Model and Optimization Algorithm in Edge Computing

With the popularity of various artificial intelligence applications in mobile devices, a large number of new demands pose new challenges to the application deployment of edge computing platforms. Due to the limited computing resources of edge devices, many AI applications cannot land because it is difficult to meet the real-time requirements. How to solve the contradiction between the algorithm requirements and computing power of edge devices has become a key issue in edge applications. In order to solve the above problems, this paper studies the deployment and optimization of deep learning algorithm on the edge computing platform, prunes the model of deep learning algorithm, and then completes the precision calibration of INT8 with Tensor RT. Finally, combining the storage management and communication optimization methods of CUDA and the integration of tensor RT layer, an inference engine suitable for practical application is obtained. Facing the demand of intelligent road sweeper project, based on the above research contents, the software realization of intelligent road garbage identification system is completed, and the deployment optimization of road garbage detection algorithm on this platform is completed according to the proposed reasoning optimization strategy. The execution speed and performance of the optimized algorithm can meet the project requirements.