In different historical periods, people have different demands for Physical Education (PE), and the educational goal of school PE will also change, and the teaching mode and requirements of school PE will also be different. Multimedia information retrieval (IR) technology refers to searching in a large-scale multimedia database according to the content and context of media and media objects. In this paper, the Back Propagation instructional level of physical training class, and model is established. The BPNN is used to analyze the Multimedia IR data, and the simulation and experiment show the effectiveness of the assessment model, which provides a new method for comprehensively, fairly and scientifically evaluating the teaching work. This method is improved by 11.85% compared with the comparative assessment algorithms. The method of using BPNN to evaluate instructional level is easy to operate, which can overcome the shortcomings of traditional assessment methods, and avoid the complicated process of traditional assessment of instructional level.