Pixel-Wise Fabric Defect Detection by CNNs Without Labeled Training Data

The Fabric Defect (FD) classification system lowers the expenses related to low-quality products by improving productivity and product quality. There are quite a number of existing fabric defect classification techniques, but most of them are not suitable for classification in a complex environment due to insufficient deep feature extraction. Fabric defect detection is difficult in the fabric industry due to complex shapes and a variety of defects. In this research, a novel clustering-based segmentation technique is proposed to segment fabric defects. Initially, the input fabric images are denoised utilizing the Wiener filter to reduce noise artifacts. Using the pre-processed images, the defective and normal regions are segmented using Fuzzy k-means clustering. The defective regions are identified and analyzed, while the normal regions are further classified using a Random Forest classifier to detect minute defects more accurately. Finally, both defective and normal fabric images are concatenated into the Fuzzy U-Net for fabric region segmentation. The proposed Fuzzy U-Net model attains a total accuracy (ACY) of 8.87%, 8.34%, 2.85%, and 4.51% better than DeepLabV3, seg net, Mask RCNN, and U-net, respectively. The proposed Clustering-based segmentation model enhances the total ACY by 30.14%, 14.65%, 55.16%, and 11.98% better than CNN, MSCDAE, FabricNet, and FN-Net, respectively. The proposed clustering-based segmentation technique yields higher accuracy than the existing models. Compared to CNNs and other deep learning models, the clustering-based segmentation significantly improves accuracy. Real-time fabric inspection is possible with it since it allows accurate defect detection in complex environments.

Automatic real-time crack detection using lightweight deep learning models

In the textile manufacturing industry, fabric defect detection becomes a necessary and essential step in quality control. The investment in this field is more than economical when reduction in labor cost and associated benefits are considered. Moreover, the development of a wholly automated inspection system requires efficient and robust algorithms. To overcome this problem, in this paper, we present a new fabric defect detection scheme which uses the local homogeneity and neural network. Its first step consists in computing a new homogeneity image denoted as H-image. The second step is devoted to the application of the discrete cosine transform (DCT) to the H-image and the extraction of different representative energy features of each DCT block. These energy features are used by the back-propagation neural network to judge the existence of fabric defect. Simulations on different fabric images and different defect aspects show that the proposed method achieves an average accuracy of 97.35%.

Matrix singular value decomposition technique is employed for the detection of defects in fabrics. Firstly, a region of interest (ROI) containing the defect is identified by a proposed adaptive partitioning technique – thus reducing the computational duty of operating over the whole image. The ROI portion of fabric image is then divided into small nonoverlapping subimages to further reduce the computational complexity and the average singular values of the subimages are calculated. To remove the interlaced warp–weft grating structure from ROI, which is the global information in the fabric image, selected singular values associated with positive average singular numbers are rejected and the fabric image is reconstructed to yield the image of the defect. Since the resulting image is saturated with noise and some unconnected parts mainly due to dissimilarity of the subimage of the fabric structure, postprocessing is carried out by binarization and edge detection to yield the edge map of the defect. Validity and feasibility of the proposed approach is established for detection of defects form images of TILDA database. The detection rate of 95% and detection success rate of 94.1% are achieved when tested over 460 samples.

Image manipulation has increased in popularity as a result of the software that is readily available for altering photos. Since the altered photographs cannot be distinguished with the human eye, they are spreading on numerous platforms, causing confusion and spreading rumours.Researchers have been working on several methods for the more accurate detection of altered photographs as a result.Better accuracy is provided by neural networks' ability to extract intricate hidden properties from images. In contrast to conventional methods of counterfeit detection, a deep learning model automatically creates the necessary features; as a result, it has emerged as the newest field of study in image forgery.In this research, we suggest an approach for detecting image forgery that is fusion-based. SqueezeNet, MobileNetV2, and ShuffleNet—three compact deep learning models—are the foundation of the decision fusion.Two phases comprise the implementation of the fusion decision system. The evaluation of the forgeries of the photos begins with the pretrained weights of the lightweight deep learning models. The outcomes of the counterfeiting of the photos are compared with the pre-trained models using the ne-tuned weights, second.In comparison to state-of-the-art techniques, the experimental results show that the fusion-based decision strategy delivers higher accuracy.The paper initially discusses various types of image forgery techniques and later on compares different approaches involving neural networks to identify forged images

This study proposes an innovative deep learning model for use in a multispectral vision system comprising a complementary metal‐oxide semiconductor image sensor and a spectrometer. To ensure accurate color recognition, the deep learning model includes an embedded adaptive automatic color temperature correction engine. By using this color temperature correction engine, the multispectral vision system can intelligently compensate for lighting and chromatic variations. To evaluate the performance of the system, we created a nine‐dimensional data set using the IT8.7/2 color target. We then used this data set to train the deep learning model. Our deep learning model outperformed other lightweight deep learning models in experiments, making it suitable for deployment on edge devices and embedded systems. We tested the ability of the multispectral vision system to classify adulterated coffee beans into their respective classes. The overall accuracy rate was more than 99.3%, indicating that out proposed multispectral vision system is effective in identifying color differences. Considering its capabilities in agricultural screening, we suggest incorporating our adaptive automatic multispectral vision system into agricultural machines for the realization of Agriculture 4.0.

PurposeIn the fabric manufacturing industry, various unfavorable factors, including machine fault and yarn breakage, can easily cause fabric defects and affect product quality, begetting huge economic losses to enterprises. Thus, automatic fabric defect detection systems have become an important development direction. Herein, the most common defects in the fabric production process, like ribbon yarn, broken yarn, cotton ball, holes, yarn shedding and stains, are detected. Current fabric defect detection systems afford low detection accuracy and a high missed detection rate for small target fabric defects. Therefore, this study proposes deep learning technology for automatically detecting fabric defects by improving the YOLOv5s target detection algorithm. The improved algorithm is termed YOLOv5s-4SCK, which can effectively detect fabric defects. This study aims to discuss the aforementioned issues.Design/methodology/approachSpecifically, based on the YOLOv5s algorithm, first, the structure of YOLOv5s is modified to add a small target detection layer, fully utilize deep and shallow features and reduce the missed detection rate of small target fabric defects. Second, the integration of CARAFE upsampling enables the effective retention of feature information and maintenance of a certain computational efficiency, thereby improving the detection accuracy. Finally, the K-Means++ clustering algorithm is used to analyze the position of the center point of the prior box to better obtain the anchor box and improve the average accuracy and evaluation index of detection.FindingsThe research results show that the YOLOv5s-4SCK algorithm increases the accuracy by 4.1% and the detection speed by 2 f.s-1 compared to the original YOLOv5s algorithm, and it effectively improves the original YOLOv5s problem of high missed detection rate of small targets.Research limitations/implicationsThe YOLOv5s-4SCK proposed in this paper can effectively reduce the missed detection rate of fabric defects, improve the detection efficiency and has certain industrial value.Practical implicationsThe proposed algorithm can quickly identify fabric defects, effectively improving the detection rate. In the future, the proposed algorithm will be applied in the actual industry.Social implicationsAutomatic fabric defect detection reduces the manpower of inspectors, and the proposed YOLOv5s-4SCK algorithm is also suitable for other recognition fields.Originality/valueThe proposed YOLOv5s-4SCK algorithm has been tested using real cloth to ensure its accuracy, and its performance is better than the original YOLOv5s algorithm.

Generative adversarial networks (GANs) have shown striking performances in computer vision applications to augment virtual training samples (VTS). However, the VTS generating by GANs in the context of hyperspectral image classification suffer from structural inconsistency due to the insufficient number of training samples in order to learn high-order features from the discriminator. This work addresses the scarcity of training samples by designing a GAN, in which the performance of discriminator is improved to produce more structurally coherent VTS. In the proposed method, by splitting the discriminator into two parts, GAN undertakes two tasks: the main task is to learn to distinguish between real and fake samples, and the auxiliary task is to learn to distinguish structurally corrupted and real samples. With this setup, GAN will produce real-like VTS with a higher variation than conventional GAN. Furthermore, in order to reduce the computational cost, subspace-based dimension reduction was performed to obtain the dominant features around the training samples to generate meaningful patterns from the original ones to be used in the learning phase. Based on the experimental results on real, and well-known hyperspectral benchmark images, the proposed method improves the performance compared with GANs-related, and conventional data augmentation strategies <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> .

To address the issues of weak defect features and high missed detection rates in infrared detection of aircraft honeycomb structures using traditional methods, this study proposes a detection scheme for infrared images of simulated honeycomb structures based on lightweight deep-learning models. Among existing deep-learning models, YOLO11-s, as a lightweight deep learning model specifically designed for industrial detection scenarios, stands out as an ideal choice for real-time detection of aviation structures due to its lightweight architecture and efficient inference performance. This research adopts the model as the core framework and integrates the advantages of mainstream deep learning models such as ResNet, VGG, and Transformer in feature extraction. It collects images of honeycomb structures with typical defects (water accumulation, debonding, and delamination) through infrared thermal imaging technology, and constructs an end-to-end defect detection model by combining adaptive histogram equalization preprocessing [1] and multiple sets of hyperparameter optimization experiments. The experimental results show that the optimized model has significantly improved detection accuracy compared with the original model, and exhibits good recognition ability for different types of defects, providing a new technical path for efficient detection of aviation structural defects.

A computer-based intelligent visual inspection system plays a major role in evaluating the quality of textile fabrics and its demand is continuously increasing in the textile industry, especially when the quality of textile is to be considered. In this paper, we propose an AI-based automated fabric defect detection algorithm which utilizes pre-trained deep neural network models for classifying possible fabric defects. The fabric images are enhanced by pre-processing at various levels using conventional image processing techniques and they are used to train the networks. The Deep Convolutional Neural Network (DCNN) and a pre-trained network, AlexNet, are used to train and classify various fabric defects. With the exiting textile dataset, a maximum classification accuracy of 92.60% is achieved in the conducted simulations. With this accuracy, the detection and classification system based on this classifier model can aid the human to find faults in the fabric manufacturing unit.

Coronary artery disease (CAD) is among the most prevalent and life-threatening cardiovascular conditions worldwide. Early detection is essential for improving patient outcomes and improving the efficiency of healthcare systems. Electrocardiography (ECG) is widely used for assessing cardiac function, yet manual interpretation of ECG signals can be inconsistent and prone to error. Developing reliable automated methods is therefore of great importance for enabling earlier and more consistent CAD detection. Although recent advances in deep learning have achieved strong performance in ECG analysis, many existing methods remain limited in practice. Current state-of-the-art models are often computationally complex and thus unsuitable for deployment on resource-constrained platforms. To address the limited attention given to CAD in current ECG research, various deep learning models were developed in this thesis. First, a one-dimensional convolutional neural network (1D-CNN) was proposed to detect CAD directly from raw ECG signals without manual feature extraction. The model achieved 97.3% accuracy and demonstrated strong generalisability when using 250-point ECG segments. Then, a feature engineering approach was applied to select high-quality signal segments using sample entropy and normalisation techniques, further improving both accuracy and robustness. Next, a lightweight neural network architecture (CADNet) was developed, outperforming existing lightweight models by offering lower complexity and smaller size without compromising accuracy. The fourth study focused on 12-lead ECG, introducing a depthwise, squeeze-and-excitation-based model that captured both lead-specific and inter-lead patterns, and achieved efficient deployment on an STM32 microcontroller. Finally, an attention-driven model was proposed for the detection of multiple cardiovascular diseases from a single ECG recording, demonstrating high diagnostic capability. Our qualitative evaluations in this study demonstrate that lightweight deep learning models can provide reliable ECG-based CAD diagnosis while remaining suitable for real-time deployment. The proposed approaches are robust to ECG variability and show consistent performance across different experimental and diagnostic scenarios, supporting their practical use in resource-constrained environments. Overall, the results highlight the relevance of lightweight deep learning architectures in enabling ECG-based diagnostic methods to be used as potential pre-screening tools.

Complete and accurate burned area map data are needed to document spatial and temporal patterns of fires, to quantify their drivers, and to assess the impacts on human and natural systems. To achieve the the purpose of identifying burned area accurately and efficiency from remote sensing images, a lightweight deep learning model is proposed based on Deeplab V3 + , which employs the combination of attention mechanism and deep transitive transfer learning (DTTL) strategy. The lightweight MobileNet V2 network integrated with Convolutional Block Attention Module (CBAM) is designed as the backbone network to replace the traditional time-consuming Xception of Deeplab V3 +. The attention mechanism is introduced to enhance the recognition ability of the proposed deep learning model, and the deep transitive transfer learning strategy is adopted to solve the problem of incorrect identification of the burned area and discontinuous edge details caused by insufficient sample size during the extraction process. For the process of DTTL, the improved Deeplab V3 + network was first pre-trained on ImageNet. Sequentially, WorldView-2 and the Sentinel-2 dataset were employed to train the proposed network based on the ImageNet pre-trained weights. Experiments were conducted to extract burned area from remote sensing images based on the trained model, and the results show that the proposed methodology can improve extraction accuracy with OA of 92.97% and Kappa of 0.819, which is higher than the comparative methods, and it can reduce the training time at the same time. We applied this methodology to identify the burned area in Western Attica region of Greece, and a satisfactory result was achieved with. OA of 93.58% and Kappa of 0.8265. This study demonstrates the effectiveness of the improved Deeplab V3 +in identifying forest burned area. which can provide valuable information for forest protection and monitoring.

Forest resources are one of the indispensable resources of the earth, which are the basis for the survival and development of human society. With the swift advancements in computer vision and artificial intelligence technology, the utilization of deep learning for smoke detection has achieved remarkable results. However, the existing deep learning models have poor performance in forest scenes and are difficult to deploy because of numerous parameters. Hence, we introduce an optimized forest fire smoke monitoring system for embedded edge devices based on a lightweight deep learning model. The model makes full use of the multi-scale variable attention mechanism of Transformer architecture to strengthen the ability of image feature extraction. Considering the needs of application scenarios, we propose an improved lightweight network model LCNet for feature extraction, which can reduce the parameters and enhance detecting ability. In order to improve running speed, a simple semi-supervised label knowledge distillation scheme is used to enhance the overall detection capability. Finally, we design and implement a forest fire smoke detection system on an embedded device, including the Jetson NX hardware platform, high-definition camera, and detection software system. The lightweight model is transplanted to the embedded edge device to achieve rapid forest fire smoke detection. Also, an asynchronous processing framework is designed to make the system highly available and robust. The improved model reduces three-fourths of the parameters and increases speed by 3.4 times with similar accuracy to the original model. This demonstrates that our system meets the precision demand and detects smoke in time.

Dermatological conditions are primarily prevalent in humans and are primarily caused by environmental and climatic fluctuations, as well as various other reasons. Timely identification is the most effective remedy to avert minor ailments from escalating into severe conditions. Diagnosing skin illnesses is consistently challenging for health practitioners. Presently, they rely on conventional methods, such as examining the condition of the skin. State-of-the-art technologies can enhance the accuracy of skin disease diagnosis by utilizing data-driven approaches. This paper presents a Computer Assisted Diagnosis (CAD) framework that has been developed to detect skin illnesses at an early stage. We suggest a computationally efficient and lightweight deep learning model that utilizes a CNN architecture. We then do thorough experiments to compare the performance of shallow and deep learning models. The CNN model under consideration consists of seven convolutional layers and has obtained an accuracy of 87.64% when applied to three distinct disease categories. The studies were conducted using the International Skin Imaging Collaboration (ISIC) dataset, which exclusively consists of dermoscopic images. This study enhances the field of skin disease diagnostics by utilizing state-of-the-art technology, attaining exceptional levels of accuracy, and striving for efficiency improvements. The unique features and future considerations of this technology create opportunities for additional advancements in the automated diagnosis of skin diseases and tailored treatment.

Many fabric defects are very small and undistinguishable, which makes them very difficult to detect by only monitoring the intensity change. Faultless fabric is a repetitive and regular global texture and Fourier transforms can be applied to monitor the spatial frequency spectrum of a fabric. When a defect occurs in fabric, its regular structure is changed so that the corresponding intensity at some specific positions of the frequency spectrum would change. However, the three-dimensional frequency spectrum is very difficult to analyze. In this paper, a simulated fabric model is used to understand the relationship between the fabric structure in the image space and in the frequency space. Based on the three-dimensional frequency spectrum, two significant spectral diagrams are defined and used for analyzing the fabric defect. These two diagrams are called the central spatial frequency spectrums. The defects are broadly classified into four classes: (1) double yarn; (2) missing yarn; (3) webs or broken fabric; and (4) yarn densities variation. After evaluating these four classes of defects using some simulated models and real samples, seven characteristic parameters for a central spatial frequency spectrum are extracted for defect classification.