Abstract: In view of the problems of the existing garbage classification methods, such as trivial trouble, low efficiency and supervision difficulty, an intelligent garbage classification system based on GoogLeNet convolutional neural network is designed to make up for the shortcomings of market products from the aspects of automatic identification and delivery, simple operation and household adaptation. Based on GoogLeNet convolutional neural network, a four-classification recognition classification algorithm based on garbage classification is designed, and the collected garbage data set is fine-tuned and trained to realize the effective recognition of garbage types. Raspberry PI as the core controller, with CSI camera to obtain images equipped with image recognition algorithm to realize garbage classification. It designed the mechanical structure and control unit combined with classification information independently to complete automatic garbage delivery. The test analysis shows that the system can accurately realize garbage type identification and delivery. It also can improve the convenience and efficiency of garbage classification effectively, and has great significance for the implementation of garbage classification on the client side.