Application of Convolutional Neural Network Based on Transfer Learning for Garbage Classification

In order to effectively utilize garbage resources, reduce environmental pollution and the burden of people sorting garbage, this paper proposes a method of garbage classification and recognition based on transfer learning, which migrates the existing InceptionV3 model recognition task on the Imagenet dataset to garbage identification. First, increase the data set through data augmentation. Then build a convolutional neural network based on the source model and adjust the neural network parameters based on the training effect. The training results show that the training accuracy is 99.3% and the test accuracy is 93.2%. Finally, the model is applied to the pictures collected in real life for recognition. The recognition results show that the model has good performance and high accuracy, can correctly identify common garbage in life, and has reference significance for intelligent garbage classification, which proves the feasibility of this method.