In semi-supervised semantic segmentation (SSSS), segmentation performance is heavily constrained by the quality of pseudo labels. However, prevalent pseudo-label optimization approaches rely on the model’s internal self-correction. When the model fails to recognize or adequately represent certain classes, this self-enhancement mechanism amplifies initial mistakes, ultimately leading to poor semantic or spatial consistency. To address this limitation, we propose ViLaDiff to enhance pseudo-label quality. Specifically, ViLaDiff first employs a prompt-guided image captioning task to generate descriptive text for each input image, providing high-level semantic context. To our knowledge, this is the first attempt to introduce vision-language modeling into SSSS. We design a vision-language fusion module to enhance feature semantics and discriminative capability. It integrates cross-modal interactions with dual-path knowledge to ensure semantic consistency. Additionally, while language provides high-level semantic guidance, it is inherently limited in expressing fine-grained spatial structures. Therefore, we propose an edge-aware mixed-noise diffusion process. It simulates feature-level uncertainty through Gaussian perturbations and introduces class-flipping noise into the masks to model misclassification errors. To enhance boundary refinement, we apply a higher flipping probability along mask edges, enabling edge-aware modeling during denoising. Extensive experiments on public benchmarks validate that our method significantly improves pseudo-label quality and segmentation performance.

Similar Morphology Cells (SMCs) play a critical role in pathology diagnosis. This study identifies and defines two primary challenges for SMCs segmentation based on detailed investigation. At the macro level of slides, the sample distribution within pathology images is highly imbalanced due to varying cell functionalities. At the micro level of cells, differences between SMCs are imperceptible. These two challenges lead to confusion between visually similar cell classes in SMCs segmentation, reducing segmentation accuracy. To address these issues, we propose Sharper Insights (SI), a novel approach designed to enhance the model’s discriminative capabilities. Specifically, we introduce a Discrimination Enhancement Module (DEM) to increase the model’s sensitivity to subtle variations among SMCs. This module utilizes two attention mechanisms to sharpen the model’s focus on minute differences. Additionally, a novel loss function, Minority Categories Augmentation (MCA), is introduced to mitigate distribution imbalance by dynamically adjusting category-specific weights. Finally, to verify the effectiveness and generalization of the proposed method SI, we curate a comprehensive breast cancer dataset SimMorphBC. It primarily focuses on SMCs, which have not been explicitly proposed in previous datasets. Extensive experiments on breast cancer pathology images demonstrate the superior performance of the proposed approach.

The domain generalization approach seeks to develop a universal model that performs well on unknown target domains with the aid of diverse source domains. Data augmentation has proven to be an effective method to enhance domain generalization in computer vision. Recently, semantic-level based data augmentation has yielded remarkable results. However, these methods focus on sampling semantic directions on feature space from intra-class and intra-domain, limiting the diversity of the source domain. To address this issue, we propose a novel approach called Inter-Class and Inter-Domain Semantic Augmentation (CDSA) for domain generalization. We first introduce a sampling-based method called CrossSmooth to obtain semantic directions from inter-class. Then, CrossVariance obtains the styles of different domains by sampling semantic directions. Our experiments on four well-known domain generalization benchmark datasets (Digits-DG, PACS, Office-Home, and DomainNet) demonstrate the effectiveness of our approach. We also validate our approach on commonly-used semantic segmentation datasets, namely GTAV, SYNTHIA, Cityscapes, Mapillary, and BDDS which also show significant improvements.

With the recent advances of the Internet of Things (IoT), innovative techniques, and concepts have emerged, such as digital twins and industrial 4.0. As one of the essential parts of a digital twin, IoT-based smart predictive maintenance (IoT-SPM) is a key enabling technology for smart manufacturing. This article introduces digital twins and their relationship to IoT-SPM and proposes a reference IoT-SPM, aiming to provide a comprehensive and systematic outlook for the IoT-SPM field. Thus, it can be used as a guide map for interested readers. To give a complete picture of the IoT-SPM ecosystem in industrial 4.0 systems, this article conducts an analysis from multiview perspectives, starting with the architecture, followed by platforms and component. The key components or requirements of an IoT-SPM ecosystem are identified and outlined, including the IoT and cyber–physical system (CPS) as the cornerstone technologies, IoT monitoring data as the base, big data platforms as the backbone, an upgraded computing paradigm as the catalyst, and machine learning-based data analysis as the main processor. This article also focuses on the issues surrounding IoT data when applying analytic models to a real-world industrial IoT system. Then, the current progresses relating IoT and IoT-SPM are depicted, and a research gap on IoT data quality is identified. In particular, regarding the identified IoT data quality problems, this article qualitatively evaluates and discusses the existing solutions. These discussions lead to several open research issues and future directions.

The Industrial Internet of Things (IIoT) enables intelligent predictive maintenance in smart manufacturing by incorporating IoT technologies, Big Data techniques, artificial intelligence, cloud computing, and other ever-developing enabling technologies. Although a large body of research has been conducted on IIoT based predictive maintenance, most work focuses on addressing only a part of the problem. However, predictive maintenance involves an ecosystem from ingesting data from sensors to displaying the results on a dashboard for engineers to visualize. With increasing requirements for real-time responses and privacy, integrating edge computing is no doubt a promising trend. This article proposes a complete and optimized IoT Big Data ecosystem embedded into a three-layer architecture for predictive maintenance applications. The proposed architecture consists of an edge layer, a cloud layer, and an application layer. The proposed edge infrastructure distributes the tasks effectively between the cloud layer and edge layer. On top of the architecture, different layers are integrated seamlessly to address reliability and scalability issues. In addition, an edge computing-assisted autoencoder is introduced and enabled by being deployed in a distributed manner to improve both performance and efficiency. For practical interest, we also provide an application programming interface-oriented implementation guideline for readers. To verify the proposed ecosystem, a real case study from industry is conducted to demonstrate the performance gain of edge computing-based Internet of Things systems in conjunction with the autoencoder-based deep learning technique.

With the rapid growth in the use of various smart digital sensors, the Internet of Things (IoT) is a swiftly growing technology, which has contributed significantly to Industry 4.0 and the promotion of IoT-based smart factories, which gives rise to the new challenges of big data analytics and the implementation of machine learning techniques. This article proposes a practical framework that combines IoT techniques, a data lake, data analysis, and cloud computing for manufacturing equipment health-state monitoring and diagnostics in smart manufacturing. It addresses all the required aspects in the realization of such a system and allows the seamless interchange of data and functionality. Due to the specific characteristics of IoT sensor data (low quality, redundant multisources, partial labeling), we not only provide a promising framework but also give detailed insights and pay considerable attention to data quality issues. In the proposed framework, an ingestion procedure is designed to manage data collection, data security, data transformation and data storage issues. To improve the quality of IoT big data, a high-noise feature filter is proposed for automated preliminary sensor selection to suppress noisy features, followed by a noisy data cleaning module to provide good quality data for unbiased diagnosis modeling. The proposed framework can achieve seamless integration between IoT big data ingestion from the physical factory and machine learning-based data analytics in the virtual systems. It is built on top of the Apache Spark processing engine, being capable of working in both big data and real-time environments. One case study has been conducted based on a four-stage syngas compressor from real industries, which won the Best Industry Application of IoT at the BigInsights Data & AI Innovation Awards. The experimental results demonstrate the effectiveness of both the proposed IoT-architecture and techniques to address the data quality issues.

The emergence of Industry 4.0 and the rapid advances in the Industrial Internet of Things (IIoT) have provided manufacturers with the ability to remotely monitor the process by deploying automatic fault detection in an IoT-based predictive maintenance system. However, the monitoring targets are now manufacturing plant-wide instead of being just a local area. Multiple types of faults are involved and the conventional centralized cloud computing-based IoT solutions always lead to a heavy burden on the network bandwidth due to the large amount of sensor data collected frequently that has to be transmitted to the central server and this leads to poor response time for the monitoring system. To address this problem, this article develops an artificial intelligence-assisted distributed system for manufacturing plant-wide predictive maintenance applications. The developed distributed system relies on the feature selection technique to identity an optimal feature subset for each type of fault and is enabled by deploying each independent model built on the obtained feature subset into different edge nodes. The distributed approach enables the data to be processed near the sensors, requiring less data to be transmitted to the central cloud server reducing network delay and delivering more accurate results. In addition, our proposed feature selection approach is especially designed to accommodate the characteristics of IIoT data such as the lack of labels. The effectiveness of the proposed method is validated using the widely used public Tennessee Eastman dataset.

With the rapid popularization and application of smart sensing devices, mobile crowd sensing (MCS) has made rapid development. MCS mobilizes personnel with various sensing devices to collect data. Task distribution as the key point and difficulty in the field of MCS has attracted wide attention from scholars. However, the current research on participant selection methods whose main goal is data quality is not deep enough. Different from most of these previous studies, this paper studies the participant selection scheme on the multitask condition in MCS. According to the tasks completed by the participants in the past, the accumulated reputation and willingness of participants are used to construct a quality of service model (QoS). On the basis of maximizing QoS, two heuristic greedy algorithms are used to solve participation; two options are proposed: task‐centric and user‐centric. The distance constraint factor, integrity constraint factor, and reputation constraint factor are introduced into our algorithms. The purpose is to select the most suitable set of participants on the premise of ensuring the QoS, as far as possible to improve the platform’s final revenue and the benefits of participants. We used a real data set and generated a simulation data set to evaluate the feasibility and effectiveness of the two algorithms. Detailedly compared our algorithms with the existing algorithms in terms of the number of participants selected, moving distance, and data quality. During the experiment, we established a step data pricing model to quantitatively compare the quality of data uploaded by participants. Experimental results show that two algorithms proposed in this paper have achieved better results in task quality than existing algorithms.

In recent years, the Internet-of-Things (IoT)-oriented smart manufacturing has become a prominent solution in realizing evolutional digital transformation. Missing data are one of the biggest problems for data preprocessing in an IoT architecture, and it is crucial that missing values are recovered to improve the reliability of monitoring applications. However, due to the high-frequency collection of sensor data, missing data in IoT bring new challenges. Several methods have been developed to recover missing IoT data by utilizing data from sensors that are geographically close to the sensor which is responsible for the missing data, or from sensors which provide data that are highly correlated to the missing data. In IoT systems, because of the transmission of a large volume of data over networks, common mode failures need to be considered where a single event can lead to the loss of data from a large number of sensors. In this situation, it would be infeasible to recover missing data from other sensors. To address this issue, in this article, we focus on missing data imputation for large gaps in univariate time-series data and propose an iterative framework using multiple segmented gap iteration called Itr-MS-STLecImp to provide the most appropriate values. The gap is first segmented into several pieces to initialize the missing value imputation process and then, we iteratively run gap reconstruction and gap concatenation to obtain the final imputation results. We validate the proposed approach using sensor data collected from real manufacturing plants in Australia and the comparison results show that the proposed Itr-MS-STLecImp outperforms the state-of-the-art methods in terms of root-mean-square error. Under different gap-length conditions, the proposed approach consistently reduces the error rate more than the baseline algorithm, and the error reduction is greater when the lengths of the gaps increase, indicating that the performance is significantly improved. These analysis results further prove the effectiveness of the multiple segmentation of missing gaps and the iteration operation.

The emergence of the Industrial Internet of Things (IIoT) to enhance manufacturing and industrial processes allows data analysts to address significant problems such as predictive maintenance. For the purpose of accurate data analysis, cleansing noisy sensor data is one of the most fundamental and necessary steps. Without first removing the noise, the anomaly detection techniques are likely to give a large number of false positives. However, using traditional outlier detection methods directly for such analysis are not appropriate as both noise and significant anomalies might exist in the sensor data. This article introduces the new challenges and proposes a novel solution to address the issue of removing noise while preserving anomalies in the IIoT Data. It proposes an approach that measures both the rate of change and deviation to compute the noise score. It employs a sliding window technique to define the analysis unit of the contrast measure which is used in conjunction with statistical techniques. Extensive experiments demonstrate that the proposed approach outperforms the other state-of-the-art noise detection methods, providing a clean data set that preserves the anomalies on which one can effectively apply anomaly detection techniques.

Artificial intelligence, big data, machine learning, cloud computing, and Internet of Things (IoT) are terms which have driven the fourth industrial revolution. The digital revolution has transformed the manufacturing industry into smart manufacturing through the development of intelligent systems. In this paper, a big data ecosystem is presented for the implementation of fault detection and diagnosis in predictive maintenance with real industrial big data gathered directly from large-scale global manufacturing plants, aiming to provide a complete architecture which could be used in industrial IoT-based smart manufacturing in an industrial 4.0 system. The proposed architecture overcomes multiple challenges including big data ingestion, integration, transformation, storage, analytics, and visualization in a real-time environment using various technologies such as the data lake, NoSQL database, Apache Spark, Apache Drill, Apache Hive, OPC Collector, and other techniques. Transformation protocols, authentication, and data encryption methods are also utilized to address data and network security issues. A MapReduce-based distributed PCA model is designed for fault detection and diagnosis. In a large-scale manufacturing system, not all kinds of failure data are accessible, and the absence of labels precludes all the supervised methods in the predictive phase. Furthermore, the proposed framework takes advantage of some of the characteristics of PCA such as its ease of implementation on Spark, its simple algorithmic structure, and its real-time processing ability. All these elements are essential for smart manufacturing in the evolution to Industry 4.0. The proposed detection system has been implemented into the real-time industrial production system in a cooperated company, running for several years, and the results successfully provide an alarm warning several days before the fault happens. A test case involving several outages in 2014 is reported and analyzed in detail during the experiment section.

Segmentation of white blood cells (WBCs) image is meaningful but challenging due to the complex internal characteristics of the cells and external factors, such as illumination and different microscopic views. This paper addresses two problems of the segmentation: WBC location and subimage segmentation. To locate WBCs, a method that uses multiple windows obtained by scoring multiscale cues to extract a rectangular region is proposed. In this manner, the location window not only covers the whole WBC completely, but also achieves adaptive adjustment. In the subimage segmentation, the subimages preprocessed from the location window with a replace procedure are taken as initialization, and the GrabCut algorithm based on dilation is iteratively run to obtain more precise results. The proposed algorithm is extensively evaluated using a CellaVision dataset as well as a more challenging Jiashan dataset. Compared with the existing methods, the proposed algorithm is not only concise, but also can produce high-quality segmentations. The results demonstrate that the proposed algorithm consistently outperforms other location and segmentation methods, yielding higher recall and better precision rates.