Defect detection of solar cell based on data augmentation

Abstract

In this paper, a true and false data fusion algorithm based on deep convolution confrontation generation network and random image Mosaic is proposed, which improves the training data volume by 800 times. At the same time, the network model is optimized with light weight to reduce model training parameters, storage space and prediction time. The experimental results show that compared with the original data set training and the model training using the traditional data enhancement algorithm, the test accuracy of the model training obtained by the fusion of real and false data is improved by 30% and 17% respectively, reaching nearly 77%. After the lightweight treatment, the model parameters were reduced to about 1/2 before the treatment, and the test time for each image was shortened from 57ms to 22ms. The research shows that the fusion algorithm can effectively help the image classification task with insufficient data to alleviate the problems of network overfitting and poor model performance. The lightweight optimization model can not only ensure the accuracy, but also compress the size of the model to speed up the testing speed. It is helpful to promote the application of intelligent algorithm in industrial production so as to save cost and improve production efficiency.