A stain defect is difficult to detect with the human eye because of its characteristic of having a very minimal difference in brightness with the local area of the surface. Recently, with the development of Deep learning, the Convolutional Neural Network (CNN) based stain defect classification method has been proposed. This paper proposes a Dual-stream CNN for stain defect classification using a Gabor filter image. Using Dual-stream structure CNN, Gabor filter images and Gray image (Original) preserve their respective features. The experiment based on the Magnetic Tile (MT) stain data set and the Compact Camera Module (CCM) stain dataset confirms that the proposed method has an improved performance based on the precision, recall, and F1-score in comparison to the Single-stream extraction-based method. Gabor filter images have an advantage in image texture analysis and can be used as an input to the CNNs. The Dual-stream structure better extracts the features needed for classification.

Anomaly detection in multivariate time series has gained significant attention in past few years. The rarity of anomalies, considerable data volatility, absence of anomaly labels, and need for real-time inference in modern applications makes it a challenging problem. A robust training mechanism is desired that can accurately detect anomalous points in dataset with significantly large sized normal data. Moreover, multivariate data recorded by multiple sensors significantly varies from normal values in case of any defect. It is crucial to detect which specific sensor’s data are abnormal for system analysis and correction. To address these issues, we present a self-supervised transformer for anomaly detection, STAD, which enables robust learning from unlabeled data by synthesizing pseudo-anomalies during training. The research contributions are manifold. Firstly, a synthetic anomaly injection method is developed for anomaly generation and self-supervised training on critical anomalous behaviors. Secondly, a multi-head context attention module is designed and embedded with Transformer architecture to map the local and global associations, effectively distinguishing rare anomalies. Lastly, we propose attentive class activation tokens mapping mechanism that facilitates anomaly attribution by tracing the influence of individual sensor inputs through the Transformer model, offering transparent and interpretable insights. Extensive experiments conducted on eight real-world datasets demonstrate that STAD consistently outperforms state-of-the-art unsupervised and semi-supervised approaches, achieving superior F1 scores and interpretability while maintaining computational efficiency suitable for real-time deployment.

A STAIN defect is difficult to detect with the naked eye because of its characteristic of having a very minimal difference in brightness with the local area of the surface. Usually, background extraction–based and Gabor filter–based texture feature–based methods have been proposed to detect STAIN defects. Recently, with the development of deep learning, the convolutional neural network (CNN)–based detection method has been proposed. Gabor filter images have an advantage in image texture analysis and can be used as an input to the CNNs. However, because the original image and the Gabor filter image have different domains, it is necessary to design a CNN that can extract features advantageous for STAIN defect classification. In this paper, a CNN for STAIN defect classification using the original image and Gabor filter image was proposed. A Gary-Level Co-occurrence Matrix (GLCM) matrix from the Gabor filter image was constructed, and then the Inverse Difference Moment (IDM) value, which is a homogeneity local feature, was extracted. Gabor filter images suitable for STAIN defect classification were selected using IDM values, and a STAIN defect detection system that uses the selected Gabor filter images and the inspection image as an input to the CNN was constructed. As a result of the experiment based on the magnetic tile data set and the compact camera module (CCM) dataset, it was confirmed that the proposed method has an improved performance based on the precision, recall, and F1-score in comparison to the single-stream CNN-based method.

This paper focuses on three-dimensional measurement system of printed circuit board components. Most automated optical inspections suffer from shadows and reflections obscured by the components. We propose a new method to restore shadow and reflection regions of components for accurate shape measurements, which defines shadow and reflection regions for restoration. The region is restored with a variational auto encoder-based restoration network. It is efficient to input information about both shadow and reflection region to restore the region. Additionally, long skip connection is added to VAE for improved accuracy. We confirmed its superiority through experiments by comparing its performance with other restoration networks such as cycle generative adversarial networks and coherent semantic attention inpainting.

This paper proposes GLASS-GNT, an anomaly detection network that synthesizes Gabor-shaped additive perturbation and filter-based anomaly data on normal images, utilizing a discriminator with a Vision Transformer. The proposed model generates various abnormal images using Gabor-shaped additive perturbation and employs filters to synthesize abnormal images that resemble normal ones. This process aims to enhance the model’s ability to detect complex anomalies. The discriminator functions as a Vision Transformer with a multi-head contextual attention module to identify outlier positions within the input image. Extensive evaluations on five real-world anomaly datasets demonstrate that the proposed method outperforms state-of-the-art techniques, achieving high image-level AUROC and proving its effectiveness in real-world applications. The code and dataset are available at: https://github.com/minho6988/GLASS-GNT