Method for improving classification marking accuracy of pattern patterns

Abstract

The invention discloses a method for improving the classification marking accuracy of a pattern graph, and the method comprises a picture preprocessing module, a classification training module and a data set cleaning and re-classification training model, the picture preprocessing module comprises the steps: equally dividing a picture into a group of sequence elements with coordinates, the classification training module comprises a plurality of self-attention, a plurality of Norm and a plurality of Feed Forward feedforward networks. The classification training module comprises a plurality of self-attention, a plurality of Norm and a plurality of Feed Forward feedforward networks. According to the method for improving the classification marking accuracy of the pattern graphs, the classification model is trained through the VIT model, the pictures with fuzzy category definition in the data set are automatically screened out through the training model, the latest iteration model is trained on the original model, and the pattern classification effect is better than that of other neural network CNN classification models.

Description

technical field [0001] The invention relates to the technical field of picture classification, in particular to a method for improving the classification and marking accuracy of pattern pictures. Background technique [0002] Image classification technology has gone through many years of development, from the first color matching, grayscale matching, to the current neural network feature extraction classification, and the neural network has also been improved by different models, mainly including these major networks Lenet, Alexnet, VGGNet , GooLeNet, ResNet, the accuracy of image classification has been greatly improved. In the pattern website, it is very important to classify and mark pictures correctly. The correct label classification can greatly improve the accuracy of the pattern recommended to users by intelligence, and can also indirectly improve the response time of the pattern search. The most important The most important thing is to allow users to find their favo...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to solve the traditional neural network classification problem in the prior art for website pattern diagrams, where the theme of the picture is not clear, the picture has many local features, there are many target subjects in the pattern diagram, and the feature field of view is distributed on the entire map. The classification effect of the method is not good, and a method to improve the classification and marking accuracy of the pattern map is proposed.

Images