Industrial appearance defect detection method based on zero sample learning, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an industrial appearance defect detection method based on zero sample learning, electronic equipment and a storage medium. The industrial appearance defect detection method based on zero sample learning comprises the following steps: collecting a picture of a to-be-detected sample; labeling the category and the attribute of the to-be-detected sample; projecting picture information of a to-be-detected sample to an attribute space according to the trained deep neural network model to obtain attribute tag information; determining the defect type of the to-be-detected sample according to the attribute label information; and according to the defect type of a to-be-detected sample, utilizing the trained regression network to locate the position of the defect. The embodiment of the invention has the following advantages: previous data and an external database can be efficiently utilized, and the dependence of the model on the number of to-be-detected defect samples is greatly reduced. According to the embodiment of the invention, the defects that the image features are not sufficiently extracted and the features are fuzzy due to fewer convolutional layers are avoided, and only five convolutional layers are used for extracting the defect features, so that the problem of huge calculation amount caused by excessive convolutional layers is avoided, the occupied resources are reduced, and the total recognition time is saved.

Description

technical field [0001] The present disclosure relates to the technical field of neural networks, in particular to a method for detecting industrial appearance defects based on zero-sample learning, electronic equipment and storage media. Background technique [0002] With the continuous development of the country's industrialization level, many items and accessories used in daily production and life are manufactured by industrial production lines. In the process of industrial production, various factors such as changes in the production environment, mechanical errors, and the quality of the blank may cause the produced products to contain a variety of defects (such as crushing, scratches, foreign matter, color, size differences, etc.), due to These defects can affect product performance or reduce user experience, so industrial appearance defect inspection methods need to detect them before leaving the factory. Although in the past ten years, the production of industrial pro...

AI Technical Summary

Problems solved by technology

[0006] However, the parameters of the deep neural network are huge, and good performance often depends on a large amount of data, which is contrary to the actual situation: the yield rate of industrial products put into mass production needs to be kept above a certain level, and the data that can be collected during production Defective products are rare and unevenly distributed

Images