Research on garbage image classification based on Vision Transformer and transfer learning
This paper takes domestic garbage images as the research object and the correct identification of domestic garbage categories as the research objective respectively in order to solve the problem of low classification accuracy and poor classification performance of small categories in the classification of domestic garbage images.The Vision Transformer model is used as the classification network architecture,and the transfer learning mechanism is used to realize the training and classification reasoning of the Vision Transformer model on the Huawei cloud garbage classification data set.The experimental results show that the classification model based on attention mechanism has higher classification accuracy than ResNet and DenseNet based on convolution structure.The classification accuracy can reach 96％,at the same time,the confusion matrix of the test set shows that the Vision Transformer classification model has high accuracy for small sample categories in the unbalanced datasets and has the application value of practical deployment and reasoning.
万方数据
维普
