检索结果-内蒙古大学图书馆

YOLO-feed: An advanced lightweight network enabling real-time , high-precision detection of feed pellets on CPU devices and its applications in quantifying individual fish feed intake

引用

AQUACULTURE 2025年 608卷

作者： Feng, Miaosheng Jiang, Pengxin Wang, Yuhang Hu, Shuimu Chen, Sijing Li, Rui Huang, Haibin Li, Ning Zhang, Boyu Ke, Qiaozhen Zhang, Yu Xu, Peng Xiamen Univ Coll Ocean & Earth Sci State Key Lab Mariculture Breeding Xiamen 361102 Peoples R China Xiamen Univ Coll Ocean & Earth Sci Fujian Key Lab Genet & Breeding Marine Organisms Xiamen 361102 Peoples R China Xiamen Univ Minist Educ Coll Ocean & Earth Sci Key Lab Underwater Acoust Commun & Marine Informat Xiamen 361102 Peoples R China Xiamen Univ Coll Ocean & Earth Sci Discipline Intelligent Instrument & Equipment Xiamen 361102 Peoples R China

With the rapid advancement of deep learning, its application in aquaculture has grown significantly. However, the widespread use of deep learning models in this field is hindered by the large number of parameters and high computational power requirements. Moreover, accurately quantifying individual feed intake is essential for genetic improvement of feed efficiency, scientific feeding practices, and increasing aquaculture yield. Unfortunately, existing methods struggle to operate in real-time on low-performance hardware. To address these challenges, this study proposes YOLO-Feed, a cost-effective and highly efficient deep learning network. YOLO-Feed significantly reduces model size by eliminating the large object detection head and its corresponding branch network from You Only Look Once (YOLO) v8s, while introducing a powerful intersection-over-union (IoU) loss to enhance performance. We trained the model on over 9000 labeled feed pellet images. The YOLO-Feed model achieved a precision of 0.966, comparable to YOLOv8s, while reducing the number of parameters to less than 1 million- a 12.3-fold compression. In comparative experiments, YOLO-Feed maintained near-optimal precision with minimal parameters and was the only model capable of processing a single image on a CPU in just 16.9 ms. Furthermore, in an indoor aquaculture experiment involving 1027 fish, YOLO-Feed was successfully applied to measure individual feed intake, achieving an average accuracy of 90.3 % over a 733-min experimental period. These results demonstrate the superior performance and scalability of the YOLO-Feed model. To expedite research on the genetic improvement of feeding efficiency, we have open - sourced this project on GitHub (https://***/miaomiaoge/YOLO-Feed).

关键词： deep learning Feed detection Individual feed intake Edge device deployment

来源：评论

学校读者我要写书评

暂无评论

SST-YOLOv5s: advancing real-time blood cell object detection through multi-headed attention mechanism

引用

SIGNAL image AND VIDEO processing 2025年第3期19卷 1-17页

作者： Zhang, Mingyu Chen, Jiaqing Xiamen Univ Wang Yanan Inst Studies Econ Xiamen 361005 Peoples R China Wuhan Univ Technol Coll Math & Stat Wuhan 430070 Peoples R China

An important method of diagnosing diseases in medicine is the examination of blood samples in which the blood cells in the blood sample are examined. To address the challenges of detecting small and variable blood cells in real-time, we introduce a small-target Swin Transformer YOLOv5s model named SST-YOLOv5s in this study. Based on YOLOv5s, the small object detection layer is added to enhance the feature extraction capability of YOLOv5s and improve the detection of small targets in the dataset. In this way, the C3 module is replaced by the Swin Transformer Blocks, which enhances the extraction of global information from the model as well as speeds up the computation of the model. Subsequently, considering the need for the testing speed of the model, Simplify Optimal Transport Assignment is used in this research to reduce the testing time of the model greatly. The experimental results demonstrate that compared with the baseline model of YOLOv5s, the SST-YOLOv5s model achieves a mAP0.5 of 93.0%, a mAP0.5:0.95 of 60.4%, and a Latency of 0.6. To the best of our knowledge, SST-YOLOv5s is the first YOLOv5s to add Swin Transformer and SimOTA to a small-target detection head and use it to detect cells and further proves that SST-YOLOv5s can be better applied in practical medical applications.

关键词： deep learning Medical image process Blood cell YOLOv5s Object detection

来源：评论

学校读者我要写书评

暂无评论

A Review Study of Intelligent Road Crack Detection: Algorithms and Systems

引用

INTERNATIONAL JOURNAL OF PAVEMENT RESEARCH AND TECHNOLOGY 2025年 1-31页

作者： Hemdan, Ezz El-Din Al-Atroush, M. E. Prince Sultan Univ Struct & Mat Res Lab Riyadh Saudi Arabia Menoufia Univ Fac Elect Engn Comp Sci & Engn Dept Menoufia Egypt Prince Sultan Univ Coll Engn Dept Engn Management Riyadh 66833 Saudi Arabia Prince Sultan Univ Engn Coll GEOTECH Res Lab 2 Riyadh Saudi Arabia

Crack and defect detection in civil engineering structures, including pavements, buildings, bridges, and roads, is vital for ensuring structural integrity, public safety, and long-term asset sustainability. While traditional detection methods have been widely used, they are often inefficient, inaccurate, and lack scalability, especially in real-time monitoring scenarios. The upward traffic volumes and growing complexity of structural systems aggravate these challenges. The key objective of this study is to discuss and review the currently utilized technologies as on crack detection and highlight the gap and future direction. Therefore, this review systematically analyzes modern developments in automated crack and defect detection. Besides, a review study was exploited to provide and discuss state-of-the-art deep learning (DL) approaches, as well as exploring the integration with advanced technologies as Internet of Things (IoT) and Digital Twins to facilitate real-time monitoring and predictive cack detection for smart road management in the context of smart cities. The study explores the limitations of traditional approaches and estimates how emerging knowledge can address problems for instance detection accuracy, workflow automation, and complex data management using Artificial Intelligence (AI) approaches as DL. The outcomes display the transformative potential of DL-based techniques when combined with IoT and Digital Twins, presenting smart crack detection, real-time data processing, and predictive potentials for smart road monitoring in the future. Also, this work explores key challenges, and future research scope, including enlightening interoperability, and enhancing real-time decision-making of crack detection for intelligent road management. This work presents a critical review for researchers and practitioners, suggesting a roadmap for leveraging advanced DL approaches for detecting pavement cracks methods to promise smart and resilient road infrastructure

关键词： Infrastructure health monitoring Civil engineering Smart road administration Crack detection Artificial intelligence Geographic information systems (GIS) and geospatial artificial intelligence (GeoAI) image processing deep learning IoT systems Digital twins

来源：评论

学校读者我要写书评

暂无评论

deep learning for real-time Satellite Pose Estimation on Tensor processing Units

引用

JOURNAL OF SPACECRAFT AND ROCKETS 2023年第3期60卷 1034-1038页

作者： Lotti, Alessandro Modenini, Dario Tortora, Paolo Saponara, Massimiliano Perino, Maria A. Univ Bologna Dept Ind Engn Alma Mater Studiorum I-47121 Forli Italy Univ di Bologna Interdept Ctr Ind Res Aerosp Alma Mater Studiorum I-47121 Forli Italy Thales Alenia Space Italia Nav & Control Dept I-10146 Turin Italy Thales Alenia Space Italia Space Econ Explorat & Int Network I-10146 Turin Italy

来源：评论

学校读者我要写书评

暂无评论

MS-DETR: a real-time multi-scale detection transformer for PCB defect detection

引用

SIGNAL image AND VIDEO processing 2025年第3期19卷 1-15页

作者： Ji, Li Huang, Chaohang Li, Haiwei Han, Wenjie Yi, Leiye Shenyang Aerosp Univ Minist Educ Coll Mech & Elect Engn Key Lab Rapid Dev & Mfg Technol Aircraft Shenyang Peoples R China Shenyang Aircraft Corp Shenyang Peoples R China

Surface defects on printed circuit boards degrade product quality and impact device performance, stability, and reliability. Accurately identifying these defects amidst a complex background is challenging due to the compact PCB layout. To address this, we propose the Multi-Scale Detection Transformer (MS-DETR), designed to enhance detection accuracy while maintaining efficiency for resource-constrained environments. MS-DETR introduces a Multi-Stage Convolution module to replace traditional convolution layers, enhancing feature extraction for small objects while reducing model complexity. The Slim-Scale Adaptive Fusion architecture, incorporating Ghost Squeeze-and-Excitation Convolution and Scale Adaptive Fusion modules, reduces network computation and enhances semantic information extraction. Additionally, integrating high-frequency and low-frequency (Hilo) information improves fine-grained feature extraction and focuses on critical features, while Inner Generalized Intersection over Union optimizes the loss function to accelerate convergence and enhance detection accuracy. Experimental results on a PCB defect dataset and a self-constructed dataset show that MS-DETR achieves a mean Average Precision of 96.90%, reducing arithmetic operations by 11.7 GFLOPs and parameter count by 5.9 M compared to real-time Detection Transformer, while maintaining a detection speed of 115 frames per second. These attributes make MS-DETR suitable for edge systems or embedded devices, balancing accuracy, computational cost, and speed effectively.

关键词： deep learning real-time detection transformer Printed circuit board Defect detection Small object detection

来源：评论

学校读者我要写书评

暂无评论

An adaptable physics-informed fault diagnosis approach via hybrid signal processing and transferable feature learning for structural/machinery health monitoring

引用

SIGNAL image AND VIDEO processing 2024年第12期18卷 9051-9066页

作者： Zarchi, Milad Shahgholi, Majid Tee, Kong Fah Shahid Rajaee Teacher Training Univ Fac Mech Engn Tehran *** Iran King Fahd Univ Petr & Minerals Dept Civil & Environm Engn Dhahran 31261 Saudi Arabia KFUPM Interdisciplinary Res Ctr Construct & Bldg Mat Dhahran 31261 Saudi Arabia

Structural damages, such as structural looseness and structural cracks, are commonly observed as the root causes of failures in industrial plants. These issues have been extensively studied, and deep diagnostic tools have shown promise in identifying and addressing them. However, these tools rely on large amounts of data, which leads to computational burdens and time consumption. To tackle this challenge, a groundbreaking technique is proposed within the context of this study. The key innovation of this approach lies in its ability to integrate information from various processing functions and utilize an efficient feature bank that facilitates the execution of an effective feature learning method based on a multisource strategy. This novel research also focuses on the selection of transferable features from multiple distributions for diagnostics involving unseen failure distributions. By minimizing the mean squared error function, which is based on various source domains, the accuracy of diagnostics is significantly improved. Furthermore, the joint minimization of diagnostics independence concerning failure distribution, as well as the dimension of the transferable feature space between the source domains, leads to enhanced diagnostics speed and feature visualization. To validate the effectiveness of this approach, a real case study of a structural/machinery vibration dataset is conducted to address the multi-fault diagnosis problem, encompassing machinery health conditions, foundation looseness, and cracks under various operational conditions. The results obtained from this study demonstrate that the proposed algorithm performs remarkably well in real diagnostics scenarios involving unseen failure distributions.

关键词： Structural/machinery health monitoring Multidomain data analysis Multisource information fusion Multiprocessing module Transfer learning Adaptable feature extraction

来源：评论

学校读者我要写书评

暂无评论

Few-shot bearing fault diagnosis using GAVMD-PWVD time-frequency image based on meta-transfer learning

引用

JOURNAL OF THE BRAZILIAN SOCIETY OF MECHANICAL SCIENCES AND ENGINEERING 2023年第5期45卷 1-16页

作者： Wei, Pengying Liu, Mingliang Wang, Xiaohang Heilongjiang Univ Dept Automat Harbin 150080 Peoples R China Key Lab Informat Fus Estimat & Detect Harbin Heilongjiang Peoples R China

Rolling bearings are crucial components in rotating machinery and often operate under high speeds and heavy loads for extended periods of time. If a bearing fails, it can disrupt the normal functioning of the machinery and lead to economic losses and even casualties. As a result, diagnosing faults in rolling bearings is critical and urgent. Currently, traditional fault diagnosis methods and deep learning-based methods are used for rolling bearing fault diagnosis. However, traditional methods require knowledge of signal processing techniques and selecting fault features through artificial algorithms. On the other hand, deep learning-based methods require a large number of labeled samples, but fault samples are often limited in practice. Additionally, there can be a problem of insufficient generalization ability when bearing working conditions change, which limits the application of deep learning in bearing fault diagnosis. To address this issue, a novel method is proposed in this paper that involves few-shot transfer learning and meta-learning. The method consists of four stages: using genetic algorithm to determine penalty factor and modal numbers adaptively in variational modal decomposition (GAVMD), combining correlation coefficient to eliminate useless modes, obtaining the instantaneous frequency characteristics of useful modes through Pseudo Wigner-Ville Distribution (PWVD), and using GAVMD with PWVD to obtain time-frequency images of the vibration signals of the rotating bearing. Finally, an improved relational network with deep coding ability and attention mechanism (AM) is constructed based on meta-transfer-learning and original relational network (MTLRN-AM). The experiments in this paper are based on the benchmark dataset of bearing fault diagnosis, and the results show that the proposed method has better multi-task learning ability in meta-learning and better classification performance in few-shot scenarios for bearing fault diagnosis. The average recogniti

关键词： Rolling bearing Fault diagnosis time-frequency image Few-shot learning Meta-learning Transfer learning Relation network

来源：评论

学校读者我要写书评

暂无评论

deep Reinforcement learning for real-time Energy Management in Smart Home

引用

IEEE SYSTEMS JOURNAL 2023年第2期17卷 2489-2499页

作者： Wei, Guixi Chi, Ming Liu, Zhi-Wei Ge, Mingfeng Li, Chaojie Liu, Xianggang Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Wuhan 430074 Peoples R China Huazhong Univ Sci & Technol Key Lab Image Proc & Intelligent Control Minist Educ Wuhan 430074 Peoples R China China Univ Geosci Sch Mech Engn & Elect Informat Wuhan 430074 Peoples R China Univ New South Wales Sch Elect Engn & Telecommun Sydney NSW 2052 Australia

Energy management in the smart home can help reduce residential energy costs by scheduling various energy consumption activities. However, accurately modeling factors, such as user behavior, renewable power generation, weather conditions, and real-time electricity prices can be challenging, making the design of an efficient energy management strategy difficult. This article proposes a real-time energy management algorithm based on deep reinforcement learning (DRL) for smart homes equipped with rooftop photovoltaics, energy storage systems, and smart appliances. The algorithm aims to minimize the energy cost while ensuring user comfort. A policy network that can output both discrete and continuous actions is designed to generate actions for different types of devices in a smart home. The proposed DRL-agent is trained using a proximal policy optimization approach with historical data and is used for real-time scheduling. Finally, simulations based on real-world data demonstrate the effectiveness and robustness of the proposed algorithm.

关键词： Smart homes HVAC Costs Home appliances real-time systems Energy management Task analysis deep reinforcement learning (DRL) energy storage system (ESS) home energy management proximal policy optimization (PPO)

来源：评论

学校读者我要写书评

暂无评论

Astronomical image Quality Assessment Based on deep learning for Resource-constrained Environments

引用

PUBLICATIONS OF THE ASTRONOMICAL SOCIETY OF THE PACIFIC 2025年第3期137卷 034502-034502页

作者： Li, Juan Zhang, Xiaoming Ge, Jiayi Bai, Chunhai Feng, Guojie Mu, Haiyang Wang, Lei Liu, Chengzhi Kang, Zhe Jiang, Xiaojun Chinese Acad Sci Natl Astron Observ CAS Key Lab Opt Astron Beijing 100101 Peoples R China Univ Chinese Acad Sci Beijing 100049 Peoples R China Chinese Acad Sci Xinjiang Astron Observ Urumqi 830011 Peoples R China Chinese Acad Sci Changchun Observ Natl Astron Observ Changchun 130117 Peoples R China

This paper presents a highly lightweight model for astronomical image quality assessment, named AQSA-Net, designed to address the challenges of evaluating image quality in scenarios with limited computational resources and rapid decision-making needs. With only 0.15B in computational cost and 0.67M parameters, AQSA-Net significantly reduces memory usage while enhancing inference speed and real-time processing. We construct a data set with eight quality categories based on actual astronomical images. We develop an efficient feature extraction and data processing method that integrates local and global image information, substantially reducing input resolution and training time. We optimize the AQSA-Net architecture and introduce a spatial attention unit, enabling the model to focus on key image areas, enhancing feature extraction while reducing computational overhead. AQSA-Net is compared with several classic deep convolutional neural networks, and experimental results show that AQSA-Net achieves state-of-the-art performance with minimal computational complexity and parameter count. Specifically, AQSA-Net achieves an accuracy of 97.63%, recall of 98.03%, precision of 97.79%, and F1-score of 97.91% on the test set. Additionally, to more accurately assess the quality of usable images, we construct a quantitative image quality factor and a quality grading system, providing quantifiable evaluation criteria for subsequent scientific research. Therefore, our method effectively distinguishes high-quality images from low-quality ones that may impact scientific projects. This provides a reliable automated quality assessment tool for large-scale, complex data sets requiring deep learning inspection. Furthermore, our image quality evaluation could support the assessment of scientific observation data.

关键词：

来源：评论

学校读者我要写书评

暂无评论

State of the Art on deep learning-enhanced Rendering Methods

引用

Machine Intelligence Research 2023年第6期20卷 799-821页

作者： Qi Wang Zhihua Zhong Yuchi Huo Hujun Bao Rui Wang State Key Laboratory of CAD&CG Zhejiang UniversityHangzhou310058China Zhejiang Laboratory Hangzhou311121China

Photorealistic rendering of the virtual world is an important and classic problem in the field of computer *** the development of GPU hardware and continuous research on computer graphics,representing and rendering virtual scenes has become easier and more ***,there are still unresolved challenges in efficiently rendering global illumination *** the same time,machine learning and computer vision provide real-world image analysis and synthesis methods,which can be exploited by computer graphics rendering *** learning-enhanced rendering combines techniques from deep learning and computer vision into the traditional graphics rendering pipeline to enhance existing rasterization or Monte Carlo integration *** state-of-the-art report summarizes recent studies of deep learning-enhanced rendering in the computer graphics ***,we focus on works of renderers represented using neural networks,whether the scene is represented by neural networks or traditional scene *** works are either for general scenes or specific scenes,which are differentiated by the need to retrain the network for new scenes.

关键词： Neural rendering computer graphics scene representation rendering post-processing

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：