Figure 1 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Source publication
On a global scale, the process of automatic defect detection represents a critical stage of quality control in textile industries. In this paper, a semantic segmentation network using a repeated pattern analysis algorithm is proposed for pixel-level detection of fabric defects, which is termed RPDNet (repeated pattern defect network). Specifically,...
Citations
... The image acquisition procedure, which is largely responsible for obtaining digital images of damaged samples, may often be carried out using a line-scan charge-coupled device (CCD) [6]. To locate and categorize the defective locations, the defect detection process is used, and occasionally, it also incorporates quantitative analysis [7]. ...
Fabric defect detection has been successfully implemented in the quality quick response system for textile manufacturing automation. It is challenging to detect fabric defects automatically because of the complexity of images and the variety of patterns in textiles. This study presented a deep learning-based IM-RCNN for sequentially identifying image defects in patterned fabrics. Firstly, the images are gathered from the HKBU database and these images are denoised using a contrast-limited adaptive histogram equalization filter to eliminate the noise artifacts. Then, the Sobel edge detection algorithm is utilized to extract pertinent attention features from the pre-processed images. Lastly, the proposed improved Mask RCNN (IM-RCNN) is used for classifying defected fabric into six classes, namely Stain, Hole, Carrying, Knot, Broken end, and Netting multiple, based on the segmented region of the fabric. The dataset that can be evaluated using the true-positive rate and false-positive rate parameters yields a higher accuracy of 0.978 for the proposed improved Mask RCNN. The proposed IM-RCNN improves the overall accuracy of 6.45%, 1.66%, 4.70%, and 3.86% better than MobileNet-2, U-Net, LeNet-5, and DenseNet, respectively.
... The conventional approach greatly depends on manual analysis for detecting FDs, which can support the functioning pipeline to rectify small defects instantly. Human analysis with eyes for FDs is the conventional technique utilized in the manufacturing sector, and visual inspections can recognize and find defects [2]. However, the human detection rate only attains up to 12 m per minute and has monotonous work with high regularity, an inefficient utility of human resources and higher costs, producing it inappropriate for utilisation in bulk manufacturing [3]. ...
The detection of fabric defects (FD) has become crucial in the fabric industry; however, it has some limitations due to the complex shapes and various types of FDs. The standard approach is to identify defects using human vision that can support workers to repair minor defects directly. However, the performance of manual detection is decreased slowly with increases in working hours. Consequently, there is a need to develop an automatic inspection system for FDs to improve fabric quality and decrease human work costs and errors. Recently, deep learning (DL) and computer vision (CV) approaches become popular for automated classification and recognition of FDs. Therefore, this study presents a Deer Hunting Optimization with a Deep Learning-Driven Automated Fabric Defect Detection and Classification (DHODL-AFDDC) method. The study aims to design and develop a hyperparameter-tuned DL approach for the automated recognition and classification of FDs. To achieve this, the DHODL-AFDDC method exploits an augmented MobileNetv3 approach for the feature extraction process. Furthermore, the efficiency of the improved MobileNetv3 approach can be boosted by the utilization of a DHO-based hyperparameter tuning process. Moreover, the recognition and classification of FDs take place by employing the bidirectional long-short-term memory (BiLSTM) technique. An extensive set of experiments were conducted to demonstrate the enhanced outcome of the DHODL-AFDDC technique. The stimulation outcomes highlighted the improved performance of the DHODL-AFDDC method compared to recent approaches.
... Therefore, there has been more interest in non-motifbased studies. The five primary categories of non-motifbased investigations are structural [1], statistical [2], model-based [3], learning-based [4] and spectral [5]. ...
... Mak et al. [8] presents a method for fabric defect detection that uses Gabor filters and morphological operations to enhance the texture and identify the defect regions. Huang and Xiang [4] developed a defect detection model based on CNN model and repeated pattern analysis. Their model uses both DeeplabV3+ and GhostNet to perform lightweight fabric defect detection. ...
Fabric defects cause both labor and raw material losses and energy costs. These undesirable situations negatively affect the competitiveness of companies in the textile sector. Traditionally, human-oriented quality control also has important limitations such as lack of attention and fatigue. Robust and efficient defect detection systems can be developed with image processing and artificial intelligence methods. This study proposes a deep learning-based method to detect and classify common fabric defects in circular knitting fabrics. The proposed method adds a fine-tuned mechanism to the MobileNetV2 deep learning model. The added fine-tuned mechanism is optimized to classify fabric defects. The proposed model has been tested on a fabric dataset containing circular knitting fabric defects. Obtained results showed that the proposed method produced desired results in fabric defect detection and classification.
... Huang and Xiang proposed a semantic partitioning network, called RPDNet, which uses an iterative model analysis algorithm for pixel-level detection of fabric defects. They used FI and TILDA datasets for model training [8]. ...
Textile production has an important share in the Turkish economy. One of the common problems in textile factories in Turkey is fabric texture defects that may occur due to textile machinery. The faulty production of the fabric adversely affects the company's economy and prestige. Many methods have been developed to achieve high accuracy in detecting defects in fabric. The aim of this study is to compare the performance of the models using the new dataset and deep learning models. The findings have determined that the Seresnet152d model, which is one of the transfer learning models, can classify with 95.38% accuracy on the generated dataset. Moreover, the majority voting gives 95.58% accuracy rate. In order to achieve high accuracy in the future, it is planned to optimize the parameters of the models used in the study with the help of swarm-oriented optimization algorithms.
... Several approaches that are based on deep learning have currently been applied for defect classification Liu and Ye, 2022), detection (Hao et al., 2021;Wan et al., 2022;Lian et al., 2019;Zhou et al., 2022;Gao et al., 2022), and segmentation (Dong et al., 2019;Tabernik et al., 2020;Huang and Xiang, 2022;Du et al., 2022;Patel et al., 2020;Sae-Ang et al., 2022;Li et al., 2022;Luo et al., 2023). Defect segmentation-based automated inspection is more general in understanding the scenes in the image and in providing important information such as defect shape, size, and severity for realtime monitoring and decision-making. ...
... Defect segmentation-based automated inspection is more general in understanding the scenes in the image and in providing important information such as defect shape, size, and severity for realtime monitoring and decision-making. Recently applied deep learningbased defect segmentation methods have mostly utilized supervised learning-based methods Huang and Xiang, 2022;Dong et al., 2019;Tabernik et al., 2020). However, the performance of the supervised method is limited by the shortage of large pixellevel labeled dataset to train the deep learning networks. ...
... Furthermore, the deep learning-based methods are efficient and can handle complex inspection processes to a level that cannot even be processed by humans. Deep learning-based algorithms have been widely applied for industrial defect classification and detection Liu and Ye, 2022;Hao et al., 2021;Wan et al., 2022;Weimer et al., 2016;Cao et al., 2023), while recently a few methods have also been applied for defect segmentation (Dong et al., 2019;Tabernik et al., 2020;Huang and Xiang, 2022;Li et al., 2022;Lin et al., 2021;Sime et al., 2023;Shi et al., 2023). Dong et al. (2019) proposed pyramid feature fusion and global context attention for steel surface defect segmentation with boundary refinement module. ...
Deep learning-based defect segmentation is one of the important tasks of machine vision in automated inspection. Supervised learning methods have recently achieved remarkable performance on this task. However, the effectiveness of the supervised methods is limited by the scarcity and high cost of pixel-level annotation of training data. Semi-supervised learning methods have been proposed for training deep learning networks using a limited amount of labeled data along with additional unlabeled data for image segmentation. Most of these methods are based on consistency regularization and pseudo labeling, where the predictions on unlabeled samples often come with noise and are unreliable, resulting in poor segmentation performance. To alleviate this problem, we propose uncertainty-aware pseudo labels, which are generated from dynamically mixed predictions of multiple decoders that leverage a shared encoder network. The estimated uncertainty guides the pseudo-label-based supervision and regularizes the training when using the unlabeled samples. In our experiments on four public datasets for defect segmentation, the proposed method outperformed the fully supervised baseline and six state-of-the-art semi-supervised segmentation methods. We also conducted an extensive ablation study to demonstrate the effectiveness of our approach in various settings. The implementation code for this work is available at https://github.com/djene-mengistu/UAPS
... Although it accurately detected the edge defects and holes, it was limited in optimizing the parameter values of the filter and was computationally intensive. Model analysis [10]-Lin et al. [11] used grayscale co-occurrence matrix and redundant contour transform to extract the segmented subimages texture features and combined it with convolutional neural network classification method to improve the recognition rate of the fabric defects; however, this method is more sensitive to lighting and image noise. These conventional detection methods require predefined thresholds to detect the presence of defects, and the extracted features must be carefully designed. ...
... where ( , ) represents the intersection ratio of the true value to the bounding box of the cluster centers, B denotes the true value, C represents the cluster center, and n denotes the total number of clustered objects. Accordingly, we selected k = 9 cluster centers, and the final generated anchor boxes are illustrated in Figure 5(c), with sizes (1, 12), (1,29), (3,10), (2,25), (1,78), (12,37), (2, 463), (9,497), and (145, 121). The size of the anchor frame was set adaptively according to the aspect ratio of the fabric defects, which improved the recognition and localization accuracy of the defect model. ...
Defects on fabric surfaces are difficult to identify owing to unsuitable computing devices, highly complex algorithms, small size, and high degree of integration with the fabric. To this end, this study proposes a lightweight fabric defect-detection network, YOLO-SCD, based on attention mechanism. The introduction of depth-wise separable convolution and the attention mechanism enhanced the capacity of the neck network to extract the defective features and increased the detection speed of the overall network. The extensive experimental results revealed that YOLO-SCD achieved an average accuracy of 82.92%, effective improvement of 8.49% in mAP, and an improvement of 37 fps compared to the original YOLOv4 on a standard fabric defect dataset. By leveraging its swift detection speed and high efficiency, YOLO-SCD excels in both the general fabric defect category and the difficult-to-detect fabric. Overall, it exhibited strong performance in detecting both minor flaws and flaws with high fabric integration. Furthermore, the proposed model was extended to steel datasets with similar characteristics.
... Owing to the developments in convolutional neural networks (CNN) over recent years, classical CNN models such as the ResNeXt [1], Dual Path Networks [2], and EfficientNet [3] have been proposed, which has been proven to possess good image recognition abilities. Consequently, it is great for researchers to continue working toward importing CNN models into inspection applications and obtain good results, such as for industrial inspection [4,5], cryo-electron tomogram classification [6], lithium-ion battery electrode defect detection [7], solar cell surface defect inspection [8], bolt joints monitoring [9] and rolling bearing robust fault diagnosis [10]. ...
Over recent years, with the advances in image recognition technology for deep learning, researchers have devoted continued efforts toward importing anomaly detection technology into the production line of automatic optical detection. Although unsupervised learning helps overcome the high costs associated with labeling, the accuracy of anomaly detection still needs to be improved. Accordingly, this paper proposes a novel deep learning model for anomaly detection to overcome this bottleneck. Leveraging a powerful pre-trained feature extractor and the skip connection, the proposed method achieves better feature extraction and image reconstructing capabilities. Results reveal that the areas under the curve (AUC) for the proposed method are higher than those of previous anomaly detection models for 16 out of 17 categories. This indicates that the proposed method can realize the most appropriate adjustments to the needs of production lines in order to maximize economic benefits.
Automatic vision-based inspection systems have played a key role in product quality assessment for decades through the segmentation, detection, and classification of defects. Historically, machine learning frameworks, based on hand-crafted feature extraction, selection, and validation, counted on a combined approach of parameterized image processing algorithms and explicated human knowledge. The outstanding performance of deep learning (DL) for vision systems, in automatically discovering a feature representation suitable for the corresponding task, has exponentially increased the number of scientific articles and commercial products aiming at industrial quality assessment. In such a context, this article reviews more than 220 relevant articles from the related literature published until February 2023, covering the recent consolidation and advances in the field of fully-automatic DL-based surface defects inspection systems, deployed in various industrial applications. The analyzed papers have been classified according to a bi-dimensional taxonomy, that considers both the specific defect recognition task and the employed learning paradigm. The dependency on large and high-quality labeled datasets and the different neural architectures employed to achieve an overall perception of both well-visible and subtle defects, through the supervision of fine or/and coarse data annotations have been assessed. The results of our analysis highlight a growing research interest in defect representation power enrichment, especially by transferring pre-trained layers to an optimized network and by explaining the network decisions to suggest trustworthy retention or rejection of the products being evaluated.
