In industrial quality inspection, large amounts of data on the desired product appearance are available at time training, while significantly few defective samples available. this study, we proposed new memory-augmented adversarial autoencoders to detect and localize defects in real-time using defect-free alone for model training. This research was conducted by reconstructing images an autoencoder detection results from Fréchet Markov distance (FMD). A threshold determined based statistical...

Traditional methods for defect detection applied in industry are complex, time-consuming, not robust and demanding professional experience due to hand-crafted features extraction pipeline design. Besides, current deep learning based general object segmentation demand a large number of region-level human annotations.Instead, we present DefectGAN weakly-supervised learning, which requires very few annotations. In practical application, images training dataset merely labeled with two...

Surface defect segmentation is a critical task in industrial quality control. Existing neural network architectures often face challenges providing both real-time performance and high accuracy, limiting their practical applicability time-sensitive, resource-constrained setting. To bridge this gap, we introduce A-Net, an A-shape lightweight specifically designed for surface segmentation. Initially, A-Net introduces pioneering A-shaped architecture tailored to efficiently handle low-level...

Surface defect detection based on computer vision remains a challenging task due to the uneven illumination, low contrast and miscellaneous patterns of defects. Current methods usually present undesirable accuracy lack adaptability for various scenes. In paper, novel illumination surface defects inspection (UISDI) method is proposed address these issues. First, multi-scale saliency (MSSD) construct coarse map obtain corresponding background regions. Second, similarity prior-based intrinsic...

The optimization of aircraft is a typical multidisciplinary and multi-objective problem. To solve this problem, the difficulty lies not only in high cost discipline performance evaluation but also complex coupling relationship between different disciplines. improve efficiency, new method proposed, including two algorithms: conditional generative adversarial nets with vector similarity (VS-CGAN) distributed single-step deep reinforcement learning transfer (TL-DSDRL). For low-cost disciplines,...

In the field of surface defect detection, there is a significant imbalance between number positive and negative samples, which has led to growing interest in positive-samples-based anomaly detection methods. Reconstruction-based methods are currently most commonly used approach, but they often struggle repair abnormal foregrounds reconstruct clear backgrounds simultaneously. To address this issue, we propose new approach called forgetting-inputting-based feature fusion multiple hierarchical...

Images in real surface defect detection scenes often suffer from uneven illumination. Retinex-based image enhancement methods can effectively eliminate the interference caused by illumination and improve visual quality of such images. However, these loss defect-discriminative information a high computational burden. To address above issues, we propose joint-prior-based (JPUIE) method. Specifically, semi-coupled retinex model is first constructed to accurately Furthermore, multiscale...

In surface defect detection, due to the extreme imbalance in number of positive and negative samples, positive-samples-based anomaly detection methods have received more attention. Specifically, reconstruction-based are most popular. However, existing either difficult repair abnormal foregrounds or reconstruct clear backgrounds. Therefore, we propose a memory-augmented auto-encoder (CMA-AE). At first, novel module (CMAM), which combines encoding memoryencoding way forgetting inputting,...

Purpose: To significantly improve dual energy CT (DECT) imaging by establishing a new theoretical framework of image‐domain material decomposition with incorporation edge‐preserving techniques. Methods: The proposed algorithm, HYPR‐NLM, combines the non‐local mean filter (NLM) HYPR‐LR (Local HighlY constrained backPRojection Reconstruction) framework. Image denoising using depends on noise level composite image which is average different images. For DECT, high‐ and low‐energy further reduce...

In the surface defect detection, there are some suspicious regions that cannot be uniquely classified as abnormal or normal. The annotating of is easily affected by factors such workers' emotional fluctuations and judgment standard, resulting in noisy labels, which turn leads to missing false detections, ultimately inconsistent judgments product quality. Unlike usual ones used for detection appear rather than mislabeled. noise occurs almost every label difficult correct evaluate. this paper,...

This paper focuses on the challenge of surface defect detection in manufacturing, particularly under conditions background variation and noise interference. To tackle this issue, a novel Background-Adaptive Surface Defect Detection Network (BANet) is proposed. The BANet enhances capabilities by improving generalization capacity through learning comparative abilities between positive samples testing samples. In order to mitigate impact three types (texture variation, translation, rotation),...