Objective: In order to improve the effect of competitive sports training, this paper applies deep learning to competitive sports training. In order to improve the feature recognition ability based on view deep learning, this paper proposes a structure-aware CNN joint learning framework to carry out effective feature fusion, and then input it to the Softmax classifier to realize the structure recognition and shape classification of the 3D model. Moreover, this paper combines the improved algorithm to construct a competitive sports training system based on deep learning. The data related to the training information integration analysis platform system is all realized through the access database management system, and the data involved in each module is managed in the form of a table in the database. Finally, this paper combines experimental analysis to analyze the performance of the system in this paper. The experimental results show that the competitive sports training system based on deep learning proposed in this paper is very effective.