DG-GAN: A High Quality Defect Image Generation Method for Defect Detection

Abstract

The surface defect detection of industrial products has become a crucial link in industrial manufacturing. It has a series of chain effects on the control of product quality, the safety of the subsequent use of products, the reputation of products, and production efficiency. However, in actual production, it is often difficult to collect defect image samples. Without a sufficient number of defect image samples, training defect detection models is difficult to achieve. In this paper, a defect image generation method DG-GAN is proposed for defect detection. Based on the idea of the progressive generative adversarial, D2 adversarial loss function, cyclic consistency loss function, a data augmentation module, and a self-attention mechanism are introduced to improve the training stability and generative ability of the network. The DG-GAN method can generate high-quality and high-diversity surface defect images. The surface defect image generated by the model can be used to train the defect detection model and improve the convergence stability and detection accuracy of the defect detection model. Validation was performed on two data sets. Compared to the previous methods, the FID score of the generated defect images was significantly reduced (mean reductions of 16.17 and 20.06, respectively). The YOLOX detection accuracy was significantly improved with the increase in generated defect images (the highest increases were 6.1% and 20.4%, respectively). Experimental results showed that the DG-GAN model is effective in surface defect detection tasks.

Image synthesis of apparel stitching defects using deep convolutional generative adversarial networks

“…Initially, fabric defect images undergo preprocessing, which includes tasks like converting them to grayscale and reducing noise. Defect detection is then accomplished by identifying and categorizing these preprocessed defect images using machine learning techniques like Support Vector Machine (SVM) or Decision Trees [ 3 ]. Zhou et al [ 19 ] utilized the projection histogram of circular regions to detect defects, employing a decision tree for defect classification.…”

“…The authors applied a min-max optimization framework to concurrently train two deep learning models: a generative model denoted as 'G' and a discriminative model referred to as 'D.' This approach is based on the concept of a zero-sum game, where the objective of one player is to maximize their gain while the other player's objective is to minimize their loss [ 3 ].…”

“…This can lead to defective products reaching the market, harms both the company's reputation and the economic gains. Thus, detecting defects is immensely important for maintaining product quality, enhancing production efficiency [ 3 ]. In the textile industry, identifying faulty sections within fabric traditionally involves a human inspection process that is both time-consuming and expensive.…”