“Significant progress has been made in the study of diffractive neural networks. Compared to traditional digital neural networks, diffractive neural networks have the advantages of high speed, low power consumption, and parallel computing, and therefore have attracted much attention. In order to explore the feasibility of using diffractive neural networks for multi task image classification and recognition, researchers designed and built a diffractive neural network system. The system adopts a spatial light modulator as the modulation of phase and amplitude weights for the diffraction neural network and the optical full connection of the network layers. It uses a CMOS camera to realize the photoelectric nonlinear activation of each diffraction layer output in the diffraction neural network and the discrimination of output image recognition results. The experimental results showed that the designed system model achieved accuracies of 94.1% and 92.1% in MNIST and Fashion MNIST image classification and recognition, respectively. The optical experiments conducted by building an optical path system also achieved satisfactory results, with accuracies of 91% and 81.7%, respectively. This study confirms that the designed diffraction neural network system can meet the needs of various image classification and recognition applications, providing new ideas for the design and construction of diffraction networks. This achievement is of great significance for promoting the development of diffraction neural networks in practical applications.”