检索结果-内蒙古大学图书馆

real-time Object Detection Using Low-Resolution Thermal Camera for Smart Ventilation Systems

IEEE ACCESS 2025年 13卷 79180-79188页

作者： Lee, Jun-Hee Kim, Eung-Tea Tech Univ Korea Dept Elect Engn Shihung 15073 South Korea

deep learning-based object detection research has primarily evolved around RGB camera imagery. While state-of-the-art models like YOLO, SSD, and RetinaNet demonstrate high accuracy in RGB images, their direct application to thermal imagery faces performance degradation due to the inherent differences between RGB and thermal characteristics, coupled with computational demands unsuitable for embedded environments. These challenges are particularly pronounced when using low-resolution thermal sensors, where object boundary ambiguity and thermal pattern uncertainty become significant obstacles. We propose FLARE (Fast and Lightweight Architecture for real-time Estimation), an ultra-lightweight deep learning model for real-time object detection in low-resolution (32 x 24) thermal images. FLARE implements a Feature Compression Block for efficient feature map compression and integration, along with a Spatial Denoise Block that enables adaptive processing across temperature regions to handle thermal sensor noise characteristics. To address the scarcity of thermal data, we developed the TOI (Thermal Object Insertion) data augmentation technique, which generates new training data while preserving thermal pattern characteristics. A composite loss function combining box, classification, and confidence losses was implemented to enhance object detection performance. Implementation on an STM32 MCU demonstrated that our proposed model achieved a higher mAP compared to existing lightweight models, while reducing Flash usage by 40.39 %, RAM by 49.02 %, inference time by 35.18 %, and MAC operations by 38.00 %. Comparative experiments with YOLOv8n, a state-of-the-art lightweight object detection model, on Raspberry Pi 4B showed that FLARE achieved higher mAP (0.72% improvement) while utilizing only 4.72% of YOLOv8n's memory consumption and achieving 11.34 times faster execution speed.

关键词： Object detection Feature extraction Computational modeling Ventilation real-time systems image coding Temperature sensors Sensor phenomena and characterization image resolution Data augmentation Computer vision deep learning embedded systems object detection thermal imaging data augmentation

来源：评论

学校读者我要写书评

暂无评论

Few-shot learning for facial expression recognition: a comprehensive survey

引用

JOURNAL OF real-time image processing 2023年第3期20卷 1-18页

作者： Kim, Chae-Lin Kim, Byung-Gyu Sookmyung Womens Univ Dept Artificial Intelligence Engn Seoul 04310 South Korea

Facial expression recognition (FER) is utilized in various fields that analyze facial expressions. FER is attracting increasing attention for its role in improving the convenience in human life. It is widely applied in human-computer interaction tasks. However, recently, FER tasks have encountered certain data and training issues. To address these issues in FER, few-shot learning (FSL) has been researched as a new approach. In this paper, we focus on analyzing FER techniques based on FSL and consider the computational complexity and processing time in these models. FSL has been researched as it can solve the problems of training with few datasets and generalizing in a wild-environmental condition. Based on our analysis, we describe certain existing challenges in the use of FSL in FER systems and suggest research directions to resolve these issues. FER using FSL can be time efficient and reduce the complexity in many other real-time processing tasks and is an important area for further research.

关键词： Human-computer interaction Facial expression recognition Few-shot learning deep learning

来源：评论

学校读者我要写书评

暂无评论

A New deep learning-Based Method for Automated Identification of Thoracic Lymph Node Stations in Endobronchial Ultrasound (EBUS): A Proof-of-Concept Study

引用

JOURNAL OF IMAGING 2025年第1期11卷 10-10页

作者： Ervik, Oyvind Rodde, Mia Hofstad, Erlend Fagertun Tveten, Ingrid Lango, Thomas Leira, Hakon O. Amundsen, Tore Sorger, Hanne Levanger Hosp Nord Trondelag Hosp Trust Clin Med N-7601 Levanger Norway Norwegian Univ Sci & Technol Fac Med & Hlth Sci Dept Circulat & Med Imaging N-7030 Trondheim Norway SINTEF Digital Dept Hlth Res N-7034 Trondheim Norway St Olavs Hosp Natl Res Ctr Minimally Invas & Image Guided Diagno N-7030 Trondheim Norway Trondheim Reg & Univ Hosp St Olavs Hosp Dept Thorac Med N-7030 Trondheim Norway

Endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) is a cornerstone in minimally invasive thoracic lymph node sampling. In lung cancer staging, precise assessment of lymph node position is crucial for clinical decision-making. This study aimed to demonstrate a new deep learning method to classify thoracic lymph nodes based on their anatomical location using EBUS images. Bronchoscopists labeled lymph node stations in real-time according to the Mountain Dressler nomenclature. EBUS images were then used to train and test a deep neural network (DNN) model, with intraoperative labels as ground truth. In total, 28,134 EBUS images were acquired from 56 patients. The model achieved an overall classification accuracy of 59.5 +/- 5.2%. The highest precision, sensitivity, and F1 score were observed in station 4L, 77.6 +/- 13.1%, 77.6 +/- 15.4%, and 77.6 +/- 15.4%, respectively. The lowest precision, sensitivity, and F1 score were observed in station 10L. The average processing and prediction time for a sequence of ten images was 0.65 +/- 0.04 s, demonstrating the feasibility of real-time applications. In conclusion, the new DNN-based model could be used to classify lymph node stations from EBUS images. The method performance was promising with a potential for clinical use.

关键词： endobronchial ultrasound AI-augmented EBUS deep learning deep neural networks lymph node classification lymph node staging

来源：评论

学校读者我要写书评

暂无评论

deep learning for Sensing Matrix Prediction in Computational Microwave Imaging With Coded-Apertures

引用

IEEE ACCESS 2024年 12卷 16844-16855页

作者： Zhang, Jiaming Sharma, Rahul Garcia-Fernandez, Maria Alvarez-Narciandi, Guillermo Abbasi, Muhammad Ali Babar Yurduseven, Okan Queens Univ Belfast Ctr Wireless Innovat CWI Sch Elect Elect Engn & Comp Sci EEECS Belfast BT3 9DT North Ireland

This work aims to simplify the characterization process of coded-apertures for computational imaging (CI) at microwave frequencies. A major benefit of the presented technique is the minimization of the processing time needed to calculate the system sensing matrix for microwave CI-based compressive sensing applications. To achieve this, a deep learning-based approach which is capable of generating the sensing matrix using features learned directly from the coded-aperture distribution is proposed. To avoid the vanishing gradient problem, the proposed deep learning network contains skip connections. Using a dataset of 1,000 testing samples, the average normalized mean-squared-error (NMSE) calculated between the sensing matrix generated by the conventional method and that predicted by the proposed network is 0.0036. Moreover, the average mean-squared-error (MSE) calculated between the images reconstructed using the conventional and the predicted sensing matrix is 0.00297. In addition to providing high-fidelity estimations with minimized error, we demonstrate that using the trained network, the prediction of the sensing matrix can be achieved in 0.212 s, corresponding to a 65% reduction in the computation time needed to calculate the sensing matrix. This has significant outcomes in achieving real-time operation of CI-based microwave imaging systems.

关键词： Computational imaging deep learning image reconstruction microwave imaging sensing matrix

来源：评论

学校读者我要写书评

暂无评论

Identification of Drowning Victims in Freshwater Bodies using Drift Prediction and image processing based on deep learning 10

Identification of Drowning Victims in Freshwater Bodies usin...

引用

10th International Conference on Advances in Computing and Communications, ICACC 2021

作者： Unnikrishnan, Anjana Roshni, A.T. Anusha, P.R. Vinny, Anju M Anuraj, C.K. Toc H Institute of Science Technology Department of Computer Science and Engineering Ernakulam India

ISBN: (纸本)9781665439190

Year after year drowning deaths are increasing tremendously, making it the 3rd leading cause of unintentional injury deaths worldwide. Drift prediction methodology is typically not used in river ecosystems and conventional methods for human rescue do not account for feasible and faster human detection. Utilization of multiple sensor data in underwater human rescue applications can capacitate faster human detection. This paper discusses the design, implementation, and testing of such an underwater human detection system, which spots the victim drifting or drowning in freshwater ecosystems. The water flow sensor attached to this portable device can calculate drift distance to track down the victim. The ultrasonic sensor activates the underwater camera upon detecting an object, to facilitate real-time human localization. We performed real-time object detection on a custom dataset by applying DarkNet-53 pre-trained weights on YOLOv3 architecture and a mean Average Precision (mAP) of 98.0% was achieved. The system attained a detection depth of 5m. Combined action of drift distance calculator and YOLOv3 real-time detection model can speed up underwater human extrication. © 2021 IEEE.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

A review of deep learning in blink detection

引用

PEERJ COMPUTER SCIENCE 2025年 11卷 e2594-e2594页

作者： Xiong, Jianbin Dai, Weikun Wang, Qi Dong, Xiangjun Ye, Baoyu Yang, Jianxiang Guangdong Polytech Normal Univ Sch Automat Guangzhou Guangdong Peoples R China Guangzhou Civil Aviat Coll Sch Aircraft Maintenance Engn Guangzhou Guangdong Peoples R China

Blink detection is a highly concerned research direction in the field of computer vision, which plays a key role in various application scenes such as human-computer interaction, fatigue detection and emotion perception. In recent years, with the rapid development of deep learning, the application of deep learning techniques for precise blink detection has emerged as a significant area of interest among researchers. Compared with traditional methods, the blink detection method based on deep learning offers superior feature learning ability and higher detection accuracy. However, the current research on blink detection based on deep learning lacks systematic summarization and comparison. Therefore, the aim of this article is to comprehensively review the research progress in deep learning-based blink detection methods and help researchers to have a clear understanding of the various approaches in this field. This article analyzes the progress made by several classical deep learning models in practical applications of eye blink detection while highlighting their respective strengths and weaknesses. Furthermore, it provides a comprehensive summary of commonly used datasets and evaluation metrics for blink detection. Finally, it discusses the challenges and future directions of deep learning for blink detection applications. Our analysis reveals that deep learning-based blink detection methods demonstrate strong performance in detection. However, they encounter several challenges, including training data imbalance, complex environment interference, real-time processing issues and application device limitations. By overcoming the challenges identified in this study, the application prospects of deep learning-based blink detection algorithms will be significantly enhanced.

关键词： Blink detection deep learning Feature extraction Transfer learning

来源：评论

学校读者我要写书评

暂无评论

deep learning and Transfer learning in Cardiology: A Review of Cardiovascular Disease Prediction Models

引用

IEEE ACCESS 2024年 12卷 193365-193386页

作者： Sunilkumar, G. Kumaresan, P. Vellore Inst Technol Sch Comp Sci Engn & Informat Syst SCORE Vellore 632014 India

Cardiovascular disorders are the primary cause of death on a global scale. The World Health Organization report indicates that approximately 18 million people die from CVD each year. Major cardiac risks include arrhythmia and coronary artery disease, among others. Recent advancements in Artificial Intelligence play a pivotal role for life-saving interventions in CVD treatment. This survey examines the latest progress in Machine learning, deep learning and Pre-trained transfer learning models for classifying and predicting CVD using a survey of 159 articles, including 36 image datasets, 41 signal data and 52 clinical data from various sources. The survey investigates cardiac risk factors, cardiac dysfunction classification, various modalities and medical image processing techniques, performance metrics and hybrid techniques. Surveys on traditional neural networks such as Convolutional Neural Networks, Artificial Neural Networks, and Recurrent Neural Networks often achieve an accuracy rate of 70% to 95%. Leveraging pre-trained architectures such as ResNet, DenseNet, AlexNet, MobileNet, EfficientNet and GoogLeNet, transfer learning models consistently outperform other approaches frequently achieving accuracy levels greater than 96%. Researchers utilize various hybrid optimization algorithms to improve overall accuracy rate. The outcome of the survey supports a precise prognosis for patients with comorbidities. The survey findings indicate limitations in the incorporation of multimodal data for real-time risk assessment. The study results possess the capacity to bridge gaps in research on cardiovascular disease prediction, thereby assisting medical practitioners in early identification and subsequent prognosis.

关键词： Diseases Arteries Cardiovascular diseases Solid modeling Biomedical imaging Surveys deep learning Transfer learning Blood flow Blood Cardiovascular disease machine learning deep CNN transfer learning optimization hybrid model

来源：评论

学校读者我要写书评

暂无评论

Hybrid features extraction for the online mineral grades determination in the flotation froth using deep learning

引用

ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE 2024年 129卷

作者： Bendaouia, Ahmed Abdelwahed, El Hassan Qassimi, Sara Boussetta, Abdelmalek Benzakour, Intissar Benhayoun, Abderrahmane Amar, Oumkeltoum Bourzeix, Francois Baina, Karim Cherkaoui, Mouhamed Hasidi, Oussama Cadi Ayyad Univ Fac Sci Semlalia Comp Sci Dept Comp Syst Engn Lab LISI Marrakech Morocco Moroccan Fdn Adv Sci Innovat & Res UM6P Rabat Morocco R&D & Engn Ctr Managem Grp Reminex Marrakech Morocco Cadi Ayyad Univ Fac Sci & Technol Comp & Syst Engn Lab L2IS Marrakech Morocco Mohammed V Univ Ecole Natl Super Informat Anal Syst ENSIAS Rabat 10000 Morocco ENSMR Engn Sch Rabat Morocco

The control of the froth flotation process in the mineral industry is a challenging task due to its multiple impacting parameters. Accurate and convenient examination of the concentrate grade is a crucial step in realizing effective and real-time control of the flotation process. The goal of this study is to employ image processing techniques and CNN-based features extraction combined with machine learning and deep learning to predict the elemental composition of minerals in the flotation froth. A real world dataset has been collected and preprocessed from a differential flotation circuit at the industrial flotation site based in Guemassa, Morocco. Using image-processing algorithms, the extracted features from the flotation froth include: the texture, the bubble size, the velocity and the color distribution. To predict the mineral concentrate grades, our study includes several supervised machine learning algorithms (ML), artificial neural networks (ANN) and convolutional neural networks (CNN). The industrial experimental evaluations revealed relevant performances with an accuracy up to 0.94. Furthermore, our proposed Hybrid method was evaluated in a real flotation process for the Zn, Pb, Fe and Cu concentrate grades, with an error of precision lesser than 4.53. These results demonstrate the significant potential of our proposed online analyzer as an artificial intelligence application in the field of complex polymetallic flotation circuits (Pb, Fe, Cu, Zn).

关键词： Machine learning deep learning Computer vision Features extraction Mining industry Industry 4.0 Flotation

来源：评论

学校读者我要写书评

暂无评论

Bangladeshi Vehicle Classification and Detection Using deep Convolutional Neural Networks With Transfer learning

引用

IEEE ACCESS 2025年 13卷 26429-26455页

作者： Md Farid, Dewan Kumer Das, Proshanta Islam, Monirul Sina, Ebna United Int Univ Dept Comp Sci & Engn Dhaka 1212 Bangladesh

Vehicle classification and detection has been a field of application for deep learning and image processing which play a very important role in intelligent transport management and AI-assisted driving systems. In this paper, we have presented a vehicle classification and detection system to detect and classify low-speed and high-speed Bangladeshi vehicles. To begin, we have implemented and tested the performance of the 11 pre-trained deep convolutional neural network (CNN) models: YOLOv8 Classify, MobileNetV2, GoogLeNet, AlexNet, ResNet-50, SqueezeNet, VGG19, DenseNet-121, Xception, InceptionV3, and NASNetMobile on the six vehicle classification and detection datasets: BIT-Vehicle, IDD, DhakaAI, Poribohon-BD, Sorokh-Poth, and VTID2. We have found that YOLOv8 Classify, MobileNetV2, and GoogLeNet models outperform other models in categorising vehicle types (e.g., car, truck, bus) in images where the vehicle is already the main subject. Next, we have customised the LabelImg image annotation tool to improve the following features: (a) Changing Label Font & Border, (b) Detecting Incorrect Labels, (c) Abbreviating Label Names, (d) Improving Crosshair & Bounding Box Guide, (e) Adding Progress Information, and (f) Improving File List Panel. We have collected data from real-world highway conditions in Dhaka city and labelled 54,556 objects from 5,460 images based on 16 Bangladeshi on-road vehicle classes. Furthermore, we have built a Bangladeshi native vehicle detection classifier for locating and identifying vehicles within larger scenes, often with multiple objects using YOLOv8 Detect and SSD-Mobilenet V2 models and later deploying this classifier into NVIDIA Jetson Nano Developer Kit (a small and powerful computer). Finally, we have tested the proposed Bangladeshi vehicle detection system with different timing, lighting, and weather conditions in several areas of Dhaka city. The proposed system can detect and classify low-speed and high-speed vehicles with an average 93%

关键词： Convolutional neural networks deep learning Vehicle detection Accuracy Urban areas Roads Transfer learning real-time systems Automobiles YOLO deep learning transfer learning vehicle classification and detection system

来源：评论

学校读者我要写书评

暂无评论

A real-time intelligent monitoring method for indoor evacuee distribution based on deep learning and spatial division

引用

JOURNAL OF BUILDING ENGINEERING 2024年 92卷

作者： Han, Litao Feng, Hengjian Liu, Guoyu Zhang, Aiguo Han, Tao Shandong Univ Sci & Technol Coll Geodesy & Geomat Qingdao Peoples R China Shandong Univ Sci & Technol Key Lab Geomat & Digital Technol Shandong Prov Qingdao Peoples R China Xiamen Univ Technol Coll Comp & Informat Engn Xiamen Peoples R China

Providing real-time and detailed indoor evacuee distribution information is of great significance for the on-site formulating and dynamic adjustment of indoor emergency evacuation strategies. However, due to limitations in hardware costs and privacy protection, existing video surveillance and other evacuee counting methods cannot achieve real-time and accurate monitoring of the number and distribution of evacuees in all indoor areas. To solve the problem, using video data from an indoor surveillance system as a data source, an intelligent monitoring method for indoor pedestrian real-time distribution based on deep learning and spatial division is proposed. Firstly, the whole indoor area is divided many subareas according to the layout of indoor space and the size of spatial units. Then, deep learning detection and tracking algorithms are used to detect and track evacuees to achieve the number of evacuees at the boundary of each subarea and their movement directions;Finally, the numbers of evacuees entering and leaving each subarea are counted to obtain the spatial distribution information of evacuees. The experimental results show that within an appropriate monitoring distance, the proposed method achieves an average F1 score of 91.85% for evacuee counting at the boundaries of each subarea, with a processing speed of 22 FPS. More comprehensive and accurate real-time monitoring of evacuee distribution in the entire indoor space can be achieved, including no covered areas of cameras. It not only helps to formulate on-site emergency evacuation in case of a fire, but also enhances the daily operation and management capabilities of buildings.

关键词： Intelligent monitoring Spatial division Evacuee distribution Building evacuation Computer vision

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：