检索结果-内蒙古大学图书馆

self-supervised learning for Temporal Action Segmentation in Industrial and Manufacturing Videos

IEEE ACCESS 2025年 13卷 39650-39665页

作者： Vybornova, Yuliya Aleshin, Maxim Illarionova, Svetlana Novikov, Ilya Shadrin, Dmitrii Nikonorov, Artem Burnaev, Evgeny Samara Natl Res Univ Mobil AI Multifunct Unmanned Aerial Syst Samara 443086 Russia Skolkovo Inst Sci & Technol Skoltech Moscow 121205 Russia Univ Sharjah Biomedically Informed Artificial Intelligence Lab BIMAI Lab Sharjah U Arab Emirates Irkutsk Natl Res Tech Univ Inst Informat Technol & Data Sci Irkutsk 664074 Russia Autonomous Nonprofit Org Artificial Intelligence R Moscow 105064 Russia

Reliable methods for process control are required in maintenance to prevent emergencies, maintain high quality of work, and minimize potential risks to workers. Strict adherence to technological processes is also an essential requirement for compliance with production standards. The development of computer vision technologies opens new prospects in the field of video analytics for manufacturing and technological processes. However, effective implementation of AI-based solutions in monitoring systems requires overcoming a number of existing challenges. Among these limitations are difficulties in collecting annotated datasets necessary for training neural network models to solve specific tasks. In this work, we propose a methodology for creating a foundation industrial model (FIM) based on deep learning technology, that allows easy adaptation of algorithms for specific tasks with minimal requirements for annotated datasets. The solution for temporal action segmentation involves the use of self-supervised learning concepts and visual transformers. In order to validate the proposed approach, a dataset was prepared from several technological processes and a part of the data was annotated to enable model training for up to 14 event categories. The tested approaches include the X3D-M architecture and ViT-L. The pre-training is performed using the V-JEPA approach. On the test subset, the proposed approach leads to the Mean over Frame (MoF) of 0.817 and 0.653 for oil change and tire replacement process recognition, respectively. The experiments carried out confirm the feasibility of the proposed approach.

关键词： Videos Monitoring Safety Data models Training Computational modeling Artificial intelligence Production Automobiles Annotations Action segmentation self-supervised learning video analytics transformers

来源：评论

学校读者我要写书评

暂无评论

self-supervised learning for Complex Activity Recognition Through Motif Identification learning

引用

IEEE TRANSACTIONS ON MOBILE COMPUTING 2025年第5期24卷 3779-3793页

作者： Xia, Qingxin Morales, Jaime Huang, Yongzhi Hara, Takahiro Wu, Kaishun Oshima, Hirotomo Fukuda, Masamitsu Namioka, Yasuo Maekawa, Takuya Hong Kong Univ Sci & Technol Guangzhou Informat Hub Guangzhou 511458 Peoples R China Osaka Univ Informat Sci & Technol Suita Osaka 5650871 Japan Toshiba Co Ltd Corp Mfg Engn Ctr Kawasaki Kanagawa 2350017 Japan

Owing to the cost of collecting labeled sensor data, self-supervised learning (SSL) methods for human activity recognition (HAR) that effectively use unlabeled data for pretraining have attracted attention. However, applying prior SSL to COMPLEX activities in real industrial settings poses challenges. Despite the consistency of work procedures, varying circumstances, such as different sizes of packages and contents in a packing process, introduce significant variability within the same activity class. In this study, we focus on sensor data corresponding to characteristic and necessary actions (sensor data motifs) in a specific activity such as a stretching packing tape action in an assembling a box activity, and propose to train a neural network in self-supervised learning so that it identifies occurrences of the characteristic actions, i.e., Motif Identification learning (MoIL). The feature extractor in the network is subsequently employed in the downstream activity recognition task, enabling accurate recognition of activities containing these characteristic actions, even with limited labeled training data. The MoIL approach was evaluated on real-world industrial activity data, encompassing the state-of-the-art SSL tasks with an improvement of up to 23.85% under limited training labels.

关键词： Activity recognition industrial domain self-supervised learning self-supervised learning wearable sensor wearable sensor wearable sensor

来源：评论

学校读者我要写书评

暂无评论

self-supervised learning for vehicle bearing fault diagnosis based on time-frequency dual-domain contrast and fusion

引用

NONLINEAR DYNAMICS 2025年第14期113卷 17385-17412页

作者： He, Deqiang Xu, Yuan Sun, Haimeng Jin, Zhenzhen Zhao, Ming Guangxi Univ Sch Mech Engn State Key Lab Featured Met Mat & Life Cycle Safety Guangxi Key Lab Mfg Syst & Adv Mfg Technol Nanning 530004 Peoples R China Guilin Univ Elect Technol Guangxi Key Lab Precis Nav Technol & Applicat Guilin 541004 Peoples R China Guangxi Liugong Machinery Co Ltd Liuzhou 545007 Peoples R China

Vehicle traction motor bearings (VTMB), as critical components of the bogie, play a vital role in guaranteeing train operational safety through condition monitoring and fault diagnosis. Nevertheless, owing to the complexity, diversity and nonlinear characteristics of VTMB fault signals and the high cost of obtaining labeled data, existing intelligent fault diagnosis techniques face significant challenges in practical applications. To address this, this paper proposes a self-supervised learning model based on time-frequency dual-domain contrast and fusion (TFDDCF). Firstly, a data augmentation module is constructed to expand the dataset size and improve the model's generalization performance. Secondly, TFDDCF mainly consists of dual encoders and a time-frequency (TF) fusion encoder. To this end, an encoder structure combining improved convolutional neural networks and a Transformer is designed to effectively extract local features in TF pairs and capture global features, enabling richer feature representations. The model is able to effectively adapt to the nonlinear dynamic characteristics of the signals and extract key fault features in the time-frequency dual domain. Subsequently, TF contrastive loss and TF fusion loss are introduced, with these two loss functions working collaboratively during pre-training to effectively align and fuse time-frequency domain features, thus learning latent fault feature representations from unlabeled data. Finally, a peak sparsity attention mechanism is proposed to reduce model complexity and enhance its ability to process long sequence data. Experimental validation was conducted using two datasets, and the results demonstrate that TFDDCF can achieve outstanding diagnostic performance in scenarios with limited labeled data. The highest diagnostic accuracy reached 99.57%, surpassing other comparison models.

关键词： Time-frequency domain self-supervised learning Fault diagnosis Encoders Vehicle bearing

来源：评论

学校读者我要写书评

暂无评论

self-supervised learning-based re-parameterization instance segmentation for hazardous object in terahertz image

引用

OPTICS AND LASERS IN ENGINEERING 2025年 184卷

作者： Guo, Zijie Luo, Shaojuan Zhao, Genping He, Chunhua Wang, Tao Wu, Heng Guangdong Univ Technol Sch Automat Guangdong Prov Key Lab Cyber Phys Syst Guangzhou 510006 Peoples R China Guangdong Univ Technol Sch Comp Guangzhou 510006 Peoples R China Guangdong Univ Technol Sch Chem Engn & Light Ind Guangzhou 510006 Peoples R China

Recently, terahertz imaging technology has shown widespread potential applications in the public security screening field. Previous terahertz hazardous object detection techniques primarily relied on manual recognition and image processing methods. Some solutions incorporating deep learning technologies depend on large quantities of high-quality data, making it challenging to achieve low-cost and high-performance detection. To solve the issue of dependency on large amounts of high-quality data, we propose a diversified dynamic structure You Only Look Once (DDS-YOLO) model based on masked image modeling and structural reparameterization for the instance segmentation of hazardous objects in terahertz security inspection images. To address the scarcity and annotating difficulty problems of terahertz security inspection image samples, we combine automatic data strategies with masked image modeling for self-supervised learning. We propose a multilevel feature refinement fusion mechanism to enhance the quality of learned feature representations. The backbone parameter transfer and fine-tuning training strategies are employed to achieve hazardous object instance segmentation on the terahertz dataset. To address the low detection accuracy issue caused by the poor terahertz image quality, we develop a reparameterizable hierarchical structure for the backbone, improve a multi-scale feature-integrating neck, and design a dynamically decoupled head with lower computational requirements to enhance the performance of the instance segmentation model. Experimental results demonstrate that the proposed model accurately outputs detection boxes, categories, and segmentation masks for hazardous objects with minimal training samples. The comparative experimental results indicate that the proposed model outperforms existing state-of-the-art methods in terms of detection performance. The proposed DDS-YOLO model achieves 59.3 % in mask mean Average Precision (mAP) and 61.3 % in box mAP, an

关键词： self-supervised learning Instance segmentation Object detection Terahertz imaging

来源：评论

学校读者我要写书评

暂无评论

self-supervised learning for Activity Recognition Based on Datasets With Imbalanced Classes

引用

IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE 2025年

作者： Hamad, Rebeen Ali Woo, Wai Lok Wei, Bo Yang, Longzhi Newcastle Univ Sch Comp Newcastle Upon Tyne NE4 5TG England Univ Northumbria Dept Comp & Informat Sci Newcastle Upon Tyne NE1 8ST England

Deep learning models have demonstrated significant efficacy across various domains, including healthcare, security, ubiquitous computing, and pervasive intelligence. Long short-term memory and convolutional neural networks, in particular, have revolutionized human activity recognition systems by effectively extracting semantic information directly from raw sensor data. However, the acquisition of large, well-annotated human activity datasets remains a considerable challenge due to privacy concerns, as well as the time and cost associated with labeling. To address this, unsupervised semantic representation learning, which bypasses the need for manual annotations, is crucial for utilizing the vast pool of unlabeled sensor data. In this study, we present a novel self-supervised learning network that extracts meaningful representations from sensor data without relying on predefined semantic labels. To address class imbalances, we enhance the synthetic minority oversampling technique, introducing an improved version called iSMOTE, which accurately labels synthetic samples. Additionally, we propose a random masking technique that removes identity mapping from input data, enabling the construction of generic semantic representations for downstream tasks. By utilizing balanced unlabeled data, our network significantly improves human activity recognition while reducing reliance on annotated data. We conducted extensive evaluations and comparisons of our proposed network against various supervised and self-supervised models using datasets from smart environments and wearable sensors. Remarkably, our network outperforms several existing methods across 12 public datasets. For instance, with just 25% of labeled data, our self-supervised network achieved an accuracy of 93.87% on the Ordonez Home A dataset, surpassing the performance of fully supervised methods such as DeepConvLSTM + Attention (84.97%), HAR + Attention (88.55%), and DCC + MSA (90.78%).

关键词： Human activity recognition self-supervised learning Data models Representation learning Monitoring Data mining Noise Computational modeling Unsupervised learning supervised learning imbalanced learning reducing annotated data self-supervised representation learning sensor data

来源：评论

学校读者我要写书评

暂无评论

self-supervised learning fabric defect segmentation using anomaly generation

引用

INTERNATIONAL JOURNAL OF CLOTHING SCIENCE AND TECHNOLOGY 2025年第2期37卷 340-354页

作者： Qi, Kankan Guo, Yanchi Zhou, Jian Jiangnan Univ Wuxi Peoples R China

PurposeThe problem of fabric defect detection is a particularly challenging task, as the fabric defects occupy only a small portion of the image pixels and it is difficult to collect sufficient samples for training deep learning-based models. The purpose of this work is to present a novel self-supervised learning method to address this ***/methodology/approachIn order to solve the problem of lack of defect samples, based on the fabric-specific degree of texture regularity, we propose an anomaly generation method to create synthetic fabric defects by destroying the normal fabric texture. To improve the detection of defects of different sizes, a global-local parallel detection mechanism is proposed. A self-supervised model including an anomaly generation module, a reconstruction subnetwork and a discriminative subnetwork is established to achieve model training without prior anomaly *** proposed method features self-supervised training, does not require no labelled anomaly data and detects anomalies by distinguishing their distance from normal samples at both global and local levels. When tested on four fabric datasets, our approach outperforms state-of-the-art unsupervised and self-supervised methods and achieves significantly higher localization ***/valueA high-fidelity fabric defects synthesis method is presented to alleviate the problem of collecting numerous fabric defects, providing a reference for other surface anomaly detection problems. The proposed global-local parallel detection mechanism can serve as a reference for other methods dedicated to detecting particular anomalies. The proposed self-supervised network model can effectively locate fabric anomalies without prior labelling information, which could be used as a framework for other model designs.

关键词： Fabric defect self-supervised learning Anomaly generation Image segmentation

来源：评论

学校读者我要写书评

暂无评论

self-supervised learning Powered by Synthetic Data From Diffusion Models: Application to X-Ray Images

引用

IEEE ACCESS 2025年 13卷 59074-59084页

作者： Hosseini, Abdullah Serag, Ahmed Weill Cornell Med Qatar AI Innovat Lab Doha Qatar

Synthetic data offers a compelling solution to the challenges associated with acquiring high-quality medical data, which is often constrained by privacy concerns and limited accessibility. This study explores the efficacy of synthetic data generated using diffusion models for training deep learning models within a self-supervised learning framework. The primary objective is to evaluate whether synthetic data can effectively preserve critical medical biomarkers and support reliable downstream tasks such as classification and segmentation. Using chest X-ray images as a case study, the results reveal that models pretrained on synthetic data achieve performance comparable to or surpassing those pretrained on real data. Specifically, in pneumonia classification task, the model trained on synthetic data outperformed established benchmarks, achieving an Area Under the Curve of 99.1 and an F1-score of 96.1%. Similarly, for segmentation tasks, the model trained on synthetic data demonstrated robust performance, attaining a Dice score of 0.85. These findings underscore a significant advancement in the generation of synthetic medical images, providing a viable approach to creating realistic, biomarker-preserving datasets that ensure patient confidentiality and enable diverse applications in medical imaging.

关键词： Synthetic data Data models Training Biological system modeling Image segmentation X-ray imaging Diffusion processes Diffusion models Biomarkers Medical diagnostic imaging Artificial intelligence biomedical imaging deep learning diffusion probabilistic self-supervised learning synthetic data

来源：评论

学校读者我要写书评

暂无评论

self-supervised learning with high-stable guidance law and label generation for USV trajectory tracking control

引用

OCEAN ENGINEERING 2025年 329卷

作者： Wang, Shaowei Yu, Wanneng Wu, Chuanbo Wang, Haibin Xiao, Longhai Chen, Yao Jimei Univ Sch Marine Engn Xiamen 361021 Peoples R China Fujian Prov Key Lab Naval Architecture & Ocean Eng Xiamen 361021 Peoples R China Fujian Engn & Res Ctr Offshore Small Green Intelli Xiamen 361021 Peoples R China Univ Strathclyde Dept Naval Architecture Ocean & Marine Engn Glasgow Scotland

Unmanned Surface Vehicle (USV) Systems confront significant challenges in achieving precise trajectory tracking, primarily attributed to their high coupling, nonlinear relationships, and external disturbances from environmental factors such as winds and currents. Addressing these obstacles is imperative for advancing the autonomy and performance of USVs. This paper introduces a self-supervised learning (SSL) based framework for USV trajectory tracking control. Firstly, we propose an adaptive look-ahead distance, which enhances the guidance law that exhibits remarkable stability, even at minimal look-ahead distances. Therefore, elevating the upper limit of guidance performance. Secondly, leveraging this refined guidance law, we develop a novel control label generation methodology specifically designed for USV trajectory tracking applications. This methodology facilitates the training of controllers via self-supervised learning, thereby circumventing the need for extensive and labor-intensive manual labeling processes. Finally, the proposed method is tested in multiple tracking scenarios, including simple and complex trajectories, and compared with the previous state-of-the-art (SOTA) approach. Simulation results demonstrate its effectiveness in achieving accurate trajectory tracking control for USVs.

关键词： USV Trajectory tracking Guidance law Label generation self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

self-supervised learning for neural topic models with variance-invariance-covariance regularization

引用

KNOWLEDGE AND INFORMATION SYSTEMS 2025年第6期67卷 5057-5075页

作者： Xu, Weiran Hirami, Kengo Eguchi, Koji Hiroshima Univ Grad Sch Adv Sci & Engn 1-4-1 Kagamiyama Higashi Hiroshima City Hiroshima 7398527 Japan

In our study, we propose a self-supervised neural topic model (NTM) that combines the power of NTMs and regularized self-supervised learning methods to improve performance. NTMs use neural networks to learn latent topics hidden behind the words in documents, enabling greater flexibility and the ability to estimate more coherent topics compared to traditional topic models. On the other hand, some self-supervised learning methods use a joint embedding architecture with two identical networks that produce similar representations for two augmented versions of the same input. Regularizations are applied to these representations to prevent collapse, which would otherwise result in the networks outputting constant or redundant representations for all inputs. Our model enhances topic quality by explicitly regularizing latent topic representations of anchor and positive samples. We also introduced an adversarial data augmentation method to replace the heuristic sampling method. We further developed several variation models including those on the basis of an NTM that incorporates contrastive learning with both positive and negative samples. Experimental results on three datasets showed that our models outperformed baselines and state-of-the-art models both quantitatively and qualitatively.

关键词： Neural topic models self-supervised learning Regularization Text analysis

来源：评论

学校读者我要写书评

暂无评论

self-supervised learning for remaining useful life prediction using simple triplet networks

引用

ADVANCED ENGINEERING INFORMATICS 2025年 64卷

作者： Liu, Chien-Liang Xiao, Bin Hsu, Shih-Sheng Natl Yang Ming Chiao Tung Univ Dept Ind Engn & Management 1001 Univ Rd Hsinchu 300 Taiwan Natl Yang Ming Chiao Tung Univ Dept Comp Sci 1001 Univ Rd Hsinchu 300 Taiwan Univ Ottawa Sch Elect Engn & Comp Sci 800 King Edward Ave Ottawa ON K1N 6N5 Canada

Remaining Useful Life (RUL) prediction is critical in optimizing predictive maintenance and resource management in industrial machinery. However, existing approaches often struggle in scenarios with limited labeled data and complex multivariate time series. To address these limitations, we propose a novel self-supervised deep learning framework utilizing a triplet network architecture, which leverages unlabeled data for robust representation learning. Our model effectively captures data similarities and parameter sharing, enabling reliable training even with scarce labeled data. Comprehensive experiments conducted on the widely used NASA-CMAPSS dataset demonstrate that our method significantly outperforms state-of-the-art models across various metrics, including RUL-Score, Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE), particularly in data-limited scenarios. Furthermore, a detailed component- wise analysis verifies the individual contributions of each element of the proposed method. These findings have important implications for industrial stakeholders, facilitating more accurate predictive maintenance strategies, improved cost management, and sustainable operations.

关键词： Remaining Useful Life (RUL) Predictive maintenance self-supervised learning Contrastive learning Time series data

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：