The enhancement of environmental protection awareness has put forward more stringent and detailed requirements for garbage classification, and the traditional classification methods have been difficult to meet the double standards of efficiency and accuracy. Therefore, the exploration and application of advanced technical means, such as the use of machine vision technology for automatic garbage classification, has become an important way to improve the classification accuracy and optimize the efficiency of resource recovery. In this design practice, an open garbage classification data set covering a variety of complex classification scenarios is successfully obtained through extensive search. This data set is not only rich in content, but also detailed in classification, which provides a solid foundation for the training of the model. Subsequently, we chose the ResNet residual network in the Pytorch-image-models library, which is excellent in the field of deep learning, as our model architecture. ResNet has been widely used in the field of image recognition because of its strong feature extraction ability and good generalization performance. By training the network specifically, we aim to enable it to accurately identify and classify all types of waste. After a series of carefully designed training processes and parameter tuning, our model finally achieved an impressive 99.12% classification accuracy. This achievement not only highlights the great potential of machine vision technology in the field of garbage classification, but also lays a solid foundation for our subsequent research and application.