Data augmentation method for light guide plate based on improved CycleGAN

Abstract:

An improved CycleGAN network method for defect data augmentation based on feature fusion and self attention residual module is proposed to address the insufficiency of defect sample data for light guide plate(LGP) in production, as well as the problem of minor defects. Two optimizations are made to the generator of CycleGAN: 1) Fusion of low resolution features obtained from partial up-sampling and down-sampling with high-resolution features. 2) Combine self attention mechanism with residual network structure to replace the original residual module. Qualitative and quantitative experiments were conducted to compare different data augmentation methods, and the results showed that the defect images of the LGP generated by the improved network were more realistic, and the accuracy of the YOLOv5 detection network for the LGP was improved by 5.6%, proving the effectiveness and accuracy of the proposed method.

Key words:Data augmentation,light guide plate,CycleGAN