A5007706109- Anomaly Detection Techniques and Applications
- Industrial Vision Systems and Defect Detection
- Fault Detection and Control Systems
- Image and Object Detection Techniques
- Image Processing Techniques and Applications
- Optical measurement and interference techniques
- Network Security and Intrusion Detection
- Non-Destructive Testing Techniques
- Cell Image Analysis Techniques
- Domain Adaptation and Few-Shot Learning
- Advanced Neural Network Applications
- Surface Roughness and Optical Measurements
- Integrated Circuits and Semiconductor Failure Analysis
- Biometric Identification and Security
- Advancements in Photolithography Techniques
- Digital Media Forensic Detection
Shandong Institute of Automation
2022-2024
Chinese Academy of Sciences
2019-2024
University of Chinese Academy of Sciences
2019-2023
Vision Technology (United States)
2023
Beijing Academy of Artificial Intelligence
2020-2023
Institute of Automation
2019-2021
In this article, an anomaly detection method based on background reconstruction is proposed to perform defect inspection the texture surface of industrial products. This consists two modules: 1) autoencoder integrated with a generative adversarial network utilized reconstruct textured original image as defect-free reference. Specifically, extra anomalous images are introduced and mapping given improve stability reconstruction. 2) A U-net trained pixel-wise analysis differences between...
Defect generation is a crucial method for solving data problems in industrial defect detection. However, the current methods suffer from of background information loss, insufficient consideration complex defects, and lack accurate annotations, which limits their application segmentation tasks. To tackle these problems, we proposed mask-guided background-preserving method, MDGAN (mask-guided adversarial networks). First, to preserve normal provide annotations generated samples, replacement...
Anomaly synthesis strategies can effectively enhance unsupervised anomaly detection. However, existing have limitations in the coverage and controllability of synthesis, particularly for weak defects that are very similar to normal regions. In this paper, we propose Global Local co-Synthesis Strategy (GLASS), a novel unified framework designed synthesize broader anomalies under manifold hypersphere distribution constraints Synthesis (GAS) at feature level (LAS) image level. Our method...
Unsupervised anomaly detection methods can identify surface defects in industrial images by leveraging only normal samples for training. Due to the risk of overfitting when learning from a single class, synthesis strategies are introduced enhance capability generating artificial anomalies. However, existing heavily rely on anomalous textures auxiliary datasets. Moreover, their limitations coverage and directionality may result failure capture useful information lead significant redundancy....
In this paper, an unsupervised defect inspection method based on anomaly detection is proposed to inspect various kinds of surface defects in the field industrial production. This consists two modules: (i) An image matching module utilized align input with a pre-specified template image. Specifically, all objects be detected will adjusted same position and angle. The aligned images can reduce difficulty training stage, facilitating subsequent feature extraction localization. (ii) After...
In the weaving of patterned fabrics, surface defects are main factor affecting quality fabrics. Due to low efficiency manual detection, a deep learning methods for fabric defect detection is introduced. Faster-RCNN target algorithm that takes into account both accuracy and speed, but original network not suitable situation where there large differences in size shape defects. Therefore, double-branch parallel small categories proposed. The structure two branches SF-Net(small branch)...
The Vision Challenge Track 1 for Data-Effificient Defect Detection requires competitors to instance segment 14 industrial inspection datasets in a data-defificient setting. This report introduces the technical details of team Aoi-overfifitting-Team this challenge. Our method focuses on key problem segmentation quality defect masks scenarios with limited training samples. Based Hybrid Task Cascade (HTC) algorithm, we connect transformer backbone (Swin-B) through composite connections inspired...
Visual surface anomaly detection focuses on the classification and location of regions that deviate from normal appearance, generally, only samples are provided for training. The reconstruction-based method is widely used, which locates anomalies by analyzing reconstruction error. However, there two problems unsettled in method. First, error sometimes large. This might mislead model to take as anomalies, named overkill problem. Second, it has been observed anomalous cannot be repaired...
Data augmentation is crucial to solve few-sample issues in industrial inspection based on deep learning. However, current data methods have not yet demonstrated on-par ability the synthesis of complex defects with pixel-level annotations. This paper proposes a new defect framework fill gap. Firstly, DCDGANc (Diversified and multi-class Controllable Defect Generation Adversarial Networks constant source images) proposed employ class labels construct inputs control category random codes...
Visual anomaly detection aims at classifying and locating the regions that deviate from normal appearance. Embedding-based methods reconstruction-based are two main approaches for this task. However, they either not efficient or precise enough industrial detection. To deal with problem, we derive POUTA (Produce Once Utilize Twice Anomaly detection), which improves both accuracy efficiency by reusing discriminant information potential in reconstructive network. We observe encoder decoder...
Personal identification has been used more and widely in information society. Iris recognition, as one of numerous methods, already applied many situations, because the iris patterns have stable, unique non-invasive features. We analyze general process recognition this paper, including image preprocessing, feature extraction, classification recognition. also study progress brought by deep learning method. It is our belief that traditional approaches play a very significant role field, while...