Improved CycleGAN for super-resolution of engineering drawings

Abstract

Abstract In intelligent manufacturing, the quality of machine translation engineering drawings directly affects manufacturing accuracy. Currently, most engineering drawing translation work is done manually, which greatly reduces production efficiency. This paper proposes an automatic translation method for welded structure engineering drawings based on the Cycle Generative Adversarial Network (CycleGAN). The CycleGAN model of unpaired transfer learning is used to learn the feature mapping of real welding engineering drawings to realize their automatic translation. U-Net and PatchGAN are used as the generator and discriminator, respectively. Based on the idea of removing the identity mapping function, a high-dimensional sparse network is proposed to replace the traditional dense network for the CycleGAN generator in order to improve noise robustness. It is found that increasing the residual block hidden layer increases the resolution of the generated graph. The improved and fine-tuned network models are experimentally validated by computing the gap between real and generated data. The results meet welding engineering precision standards and solve the problem of low drawing recognition efficiency in the welding manufacturing process. After training with our model, the peak signal-to-noise ratio, structural similarity and mean squared error of welding engineering drawings reach about 44.89%, 99.58% and 2.11, respectively, which are superior to the results achieved using traditional networks in terms of both training speed and accuracy.