检索结果-内蒙古大学图书馆

CASTNet: A Context-aware, Spatio-temporal Dynamic Motion Prediction Ensemble for Autonomous Driving

ACM TRANSACTIONS ON CYBER-PHYSICAL SYSTEMS 2024年第2期8卷 1-20页

作者： Mortlock, Trier Malawade, Arnav Vaibhav Tsujio, Kohei Al Faruque, Mohammad Abdullah Univ Calif Irvine 5440 Engn Hall Irvine CA 92697 USA

Autonomous vehicles are cyber-physical systems that combine embedded computing and deep learning with physical systems to perceive the world, predict future states, and safely control the vehicle through changing environments. The ability of an autonomous vehicle to accurately predict the motion of other road users across a wide range of diverse scenarios is critical for both motion planning and safety. However, existing motion prediction methods do not explicitly model contextual information about the environment, which can cause significant variations in performance across diverse driving scenarios. To address this limitation, we propose CASTNet: a dynamic, context-aware approach for motion prediction that (i) identifies the current driving context using a spatio-temporal model, (ii) adapts an ensemble of motion prediction models to fit the current context, and (iii) applies novel trajectory fusion methods to combine predictions output by the ensemble. This approach enables CASTNet to improve robustness by minimizing motion prediction error across diverse driving scenarios. CASTNet is highly modular and can be used with various existing image processing backbones and motion predictors. We demonstrate how CASTNet can improve both CNN-based and graph-learning-based motion prediction approaches and conduct ablation studies on the performance, latency, and model size for various ensemble architecture choices. In addition, we propose and evaluate several attention-based spatio-temporal models for context identification and ensemble selection. We also propose a modular trajectory fusion algorithm that effectively filters, clusters, and fuses the predicted trajectories output by the ensemble. On the nuScenes dataset, our approach demonstrates more robust and consistent performance across diverse, real-world driving contexts than state-of-the-art techniques.

关键词： Autonomous vehicles information fusion dynamic machine learning architectures context-aware models ensemble learning motion planning

来源：评论

学校读者我要写书评

暂无评论

AI-based robotic trap for real-time insect detection, monitoring and population prediction 6

AI-based robotic trap for real-time insect detection, monito...

引用

6th IEEE International Conference on image processing, Applications and Systems, IPAS 2025

作者： Kapetas, Dimitrios Christakakis, Panagiotis Goglia, Valerio Pechlivani, Eleftheria Maria Centre for Research and Tehcnology Hellas Information Technologies Institute Thessaloniki Greece Agrorobotica Srl Italy

ISBN: (纸本)9798331506520

timely, accurate, and efficient pest detection is essential in agricultural sectors to mitigate economic losses and environmental impact, mainly in insect spreads. Today, AI-based robotic traps are used for insect monitoring and forecasting population growth, enabling rapid responses in pest management to reduce pesticide use and achieve sustainable goals. The core of this challenge is to build fast and accurate Artificial Intelligence (AI) models that promptly alert farmers for immediate action. In this study, we demonstrate the innovative potential of a robust AI-based trap using a robust dataset created in real settings consisting of 217 images and 6787 total annotations. The SpyFly robotic trap was used for continuous real-time monitoring and detection of the invasive insect Tuta absoluta. The detection system leverages deep learning (DL) models, with the best-performing model (YOLOv8x) achieving high accuracy, reaching a mean Average Precision (mAP50) of 97.6%. Additionally, environmental data such as temperature, humidity, and barometric pressure captured by the trap are used to forecast insect population growth utilizing various Machine learning (ML) models. The ARIMAX model demonstrates significant accuracy in anticipating pest outbreaks, with a Mean Square Error (MSE) of 36.69 for weekly averaged predictions and 309.67 for daily predictions. This translates to an offset of 6 insects for weekly predictions and 17 insects for daily predictions, indicating the average deviation between the predicted and actual insect counts. This integrated approach supports early intervention and sustainable pest management, highlighting the potential of AI and DL in smart agriculture enhancing productivity and reducing pesticide usage. Dataset is available for download at https://***/10.5281/zenodo.13134110. © 2025 IEEE.

关键词： deep learning digital tools insect monitoring insect population forecasting pest detection pesticide reduction robotic traps smart agriculture Tuta absoluta

来源：评论

学校读者我要写书评

暂无评论

Classification of Different Stages of Eczema Disease using YOLOv8 8

Classification of Different Stages of Eczema Disease using Y...

引用

8th International Conference on Electronics, Communication and Aerospace Technology, ICECA 2024

作者： Ezhilin Freeda, S. Hariharan, T. Dinesh, R. Muthuaravind, M. Sri Ramakrishna Engineering College Department of Computer Science and Engineering Coimbatore India

ISBN: (纸本)9798350367904

Skin diseases pose significant challenges in both diagnosis and treatment due to their diverse manifestations and complexities. In recent years, the advancement of deep learning techniques has revolutionized medical image analysis, offering promising avenues for automated detection and categorization of dermatological conditions. The primary focus of this system is the detection of different stages of the skin disease 'Eczema' using YOLO model and will be integrated with mobile application. Our methodology entails training the YOLOv8 model on a dataset encompassing 3000 diverse instances of Eczema at various stages. The model undergoes fine-tuning to discern distinctive visual patterns corresponding to different phases of the skin condition. Subsequently, the trained YOLOv8 model is deployed for real-time inference on new images, enabling prompt and reliable identification of Eczema stages. Post-collection, the dataset undergoes labelling to categorically define the various stages of eczema. Performance evaluation of this model demonstrates a commendable accuracy rate of approximately 81.9%, reflecting its robustness in identifying different stages of Eczema. © 2024 IEEE.

关键词： deep learning

来源：评论

学校读者我要写书评

暂无评论

Integrated Violence and Weapon Detection Using deep learning 1

Integrated Violence and Weapon Detection Using Deep Learning

引用

1st International Conference on Pioneering Developments in Computer Science and Digital Technologies, IC2SDT 2024

作者： Yadav, Vaibhav Kumar, Sanskar Goyal, Atul Bhatla, Sidharth Sikka, Geeta Kaur, Amandeep National Institute of Technology Department of CSE Delhi India

ISBN: (纸本)9798350365016

Detecting violence in public spaces is crucial for ensuring safety and security. While current methods primarily rely on body movement cues, they often struggle to identify instances like single-gunshot violence. In this study, an innovative approach is proposed to overcome this limitation by integrating weapon detection to enhance accuracy and reduce false positives. Drawing inspiration from recent advancements in deep learning, techniques such as feature extraction using pre-trained networks are leveraged. Our methodology involves the combination of convolutional neural networks (CNNs) for spatial feature extraction and long short-term memory (LSTM) networks for temporal relation learning. Additionally, the efficacy of integrating transformer-based models, such as the Vision Transformer, for enhanced feature representation is explored. Through comprehensive evaluation on diverse datasets, our approach demonstrates high accuracy while maintaining real-time processing speed. Comparative analysis against state-of-the-art methods reveals superior performance, achieving an accuracy of up to 98% and a processing speed of 131 frames/sec. This research significantly contributes to the enhancement of violence detection systems by addressing challenges related to accuracy, speed, and applicability across various video sources. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

VIVAR: learning View-Invariant Embedding for Video Action Recognition 8

VIVAR: Learning View-Invariant Embedding for Video Action Re...

引用

8th International Conference on Video and image processing, ICVIP 2024

作者： Hasan, Zahid Ahmed, Masud Md Faridee, Abu Zaher Purushotham, Sanjay Lee, Hyungtae Kwon, Heesung Roy, Nirmalya University of Maryland Baltimore County Baltimore United States Amazon United States BlueHalo United States United States

ISBN: (数字)9781510689244

ISBN: (纸本)9781510689237

deep learning has achieved state-of-the-art video action recognition (VAR) performance by comprehending action-related features from raw video. However, these models often learn to jointly encode auxiliary view (viewpoints and sensor properties) information with primary action features, leading to performance degradation under novel views and security concerns by revealing sensor types and locations. Here, we systematically study these shortcomings of VAR models and develop a novel approach, VIVAR, to learn view-invariant spatiotemporal action features removing view information. In particular, we leverage contrastive learning to separate actions and jointly optimize adversarial loss that aligns view distributions to remove auxiliary view information in the deep embedding space using the unlabeled synchronous multiview (MV) video to learn view-invariant VAR system. We evaluate VIVAR using our in-house large-scale time synchronous MV video dataset containing 10 actions with three angular viewpoints and sensors in diverse environments. VIVAR successfully captures view-invariant action features, improves inter and intra-action clusters’ quality, and outperforms SoTA models consistently with 8% more accuracy. We additionally perform extensive studies with our datasets, model architectures, multiple contrastive learning, and view distribution alignments to provide VIVAR insights. We open-source our code and dataset to facilitate further research in view-invariant systems. © 2025 SPIE.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

Rough-to-Fine Model-based Non-contact Heart Rate Estimation using MIMO FMCW Radar

Rough-to-Fine Model-based Non-contact Heart Rate Estimation ...

引用

2024 IEEE Global Communications Conference, GLOBECOM 2024

作者： Shi, Xintong Bouazizi, Mondher Ohtsuki, Tomoaki Keio University Graduate School of Science and Technology Japan Keio University Department of Information and Computer Science Japan

ISBN: (纸本)9798350351255

MIMO FMCW Radar, a non-contact heart rate (HR) monitoring technology, has emerged as one of the superior alternatives to contact-based sensors, providing precise HR estimation without the drawbacks of discomfort or privacy concerns, and excelling in diverse environmental conditions. Recent HR estimation studies using conventional methods have achieved high accuracy but face challenges with lengthy processing times and sensitivity to experimental conditions, affecting real-time application and robustness. A previous deep learning (DL)-based approach improves on these aspects but is limited by the range resolution of SISO FMCW Radar and the requirement for close positioning of subjects. Additionally, this method has inability to utilize contextual information from adjacent time windows restricts its effectiveness for time series analysis. Thus, our proposed method combines a Curve-Length (CL) approach with a DL rough-to-fine model, addressing the limitations of previous studies and improving HR estimation accuracy and robustness. This integrated process begins with human location detection through a CL-based approach and is followed by a two-phase rough-to-fine HR estimation. The experimental results show significant improvement over the conventional methods. © 2024 IEEE.

关键词： deep learning heart rate estimation MIMO FMCW radar rough-to-fine model

来源：评论

学校读者我要写书评

暂无评论

EA-YOLO: An Efficient and Accurate UAV image Object Detection Algorithm

引用

IEEJ TRANSACTIONS ON ELECTRICAL AND ELECTRONIC ENGINEERING 2025年第1期20卷 61-68页

作者： Dong, Dehao Li, Jianzhuang Liu, Haiying Deng, Lixia Gu, Jason Liu, Lida Li, Shuang Qilu Univ Technol Shandong Acad Sci Sch Informat & Automat Engn Jinan 250353 Peoples R China Dalhousie Univ Sch Elect & Comp Engn Halifax NS B3J 1Z1 Canada Shandong Runyi Intelligent Technol Co Ltd Jinan 250002 Peoples R China

An improved EA-YOLO object detection algorithm based on YOLOv5 is proposed to address the issues of drastic changes in target scale, low detection accuracy, and high miss rate in unmanned aerial vehicle aerial photography scenarios. Firstly, a DFE module was proposed to improve the effectiveness of feature extraction and enhance the whole model's ability to learn residual features. Secondly, a CWFF architecture was introduced to enable deeper feature fusion and improve the effectiveness of feature fusion. Finally, in order to solve the traditional algorithm's shortcomings it is difficult to detect small targets. We have designed a novel SDS structure and adopted a strategy of reusing low-level feature maps to enhance the network's ability to detect small targets, making it more suitable for detecting some small objects in drone images. Experiments in the VisDrone2019 dataset demonstrated that the proposed EA-YOLOs achieved an average accuracy mAP@0.5 of 39.9%, which is an 8% improvement over YOLOv5s, and mAP@0.5:0.95 of 22.2%, which is 5.2% improvement over the original algorithm. Compared with YOLOv3, YOLOv5l, and YOLOv8s, the mAP@0.5 of EA-YOLOs improved by 0.9%, 1.8%, and 0.6%, while the GFLOPs decreased by 86.4%, 80.6%, and 26.7%. (c) 2024 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

关键词： YOLOv5 small object detection UAV image real-time detection deep learning

来源：评论

学校读者我要写书评

暂无评论

Remote sensing image scene recognition based on densenet-169

Remote sensing image scene recognition based on densenet-169

引用

2024 International Conference on image, Signal processing, and Pattern Recognition, ISPP 2024

作者： Cui, Xiyue School of Information Science and Technology Dalian Maritime University Dalian116026 China

ISBN: (纸本)9781510680425

Scene classification has become an effective technique for classifying high spatial resolution remote sensing ***, in the traditional deep learning convolutional neural network, as the remote sensing image passes through the layers in the network, some of the features will be gradually lost, resulting in a significant decrease in the classification accuracy and precision of scene recognition, and there is the problem of underutilization of *** addition, remote sensing images themselves have high *** overcome these challenges, we adopt the DenseNet ***, we first train remote sensing images from the UCMerced dataset as network ***, we introduced the DenseNet-169 model based on migration *** with DenseNet-121, DenseNet-169 has more neural network layers, and this difference is mainly manifested in the number of convolutional layers in the dense ***-169 has more convolutional layers, which increases the complexity and the number of parameters of the model, bringing the following advantages: stronger expressive power, which enables the extraction of more complex feature patterns;faster training time, thanks to the densely-connected nature, which efficiently utilizes the gradient flow;and better generalization ability, especially on large-scale complex *** our experiments, the introduced DenseNet-169 shows excellent performance compared to other state-of-the-art deep convolutional networks on the UCMerced dataset, with an accuracy of 95.14%, a precision of 95.31%, a Kappa coefficient of 94.90%, and an F1-score of 95.11%.The experimental results show that the method can make full use of the features of remote sensing images and show good visual effect, providing a good method for scene recognition. © 2024 SPIE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Novel hybrid transfer neural network for wheat crop growth stages recognition using field images

引用

SCIENTIFIC REPORTS 2025年第1期15卷 1-16页

作者： Naseer, Aisha Amjad, Madiha Raza, Ali Munir, Kashif Smerat, Aseel Gongora, Henry Fabian Rios, Carlos Eduardo Uc Ashraf, Imran Khwaja Fareed Univ Engn & Informat Technol Inst Informat Technol Rahim Yar Khan 64200 Pakistan Univ Lahore Dept Software Engn Lahore 54000 Pakistan Chitkara Univ Chitkara Univ Inst Engn & Technol Ctr Res Impact & Outcome Rajpura 140401 India Mazaya Univ Coll Comp Technol Engn Nasiriyah Iraq Al Ahliyya Amman Univ Fac Educ Sci Amman 19328 Jordan Univ Europea Atlantico Isabel Torres 21 Santander 39011 Spain Univ Int Iberoamer UNINI Campeche 24560 Mexico Univ La Romana La Romana Dominican Rep Univ Int Iberoamer Arecibo Arecibo PR 00613 USA Yeungnam Univ Dept Informat & Commun Engn Gyongsan 38541 South Korea

Wheat is one of the world's most widely cultivated cereal crops and is a primary food source for a significant portion of the population. Wheat goes through several distinct developmental phases, and accurately identifying these stages is essential for precision farming. Determining wheat growth stages accurately is crucial for increasing the efficiency of agricultural yield in wheat farming. Preliminary research identified obstacles in distinguishing between these stages, negatively impacting crop yields. To address this, this study introduces an innovative approach, MobDenNet, based on data collection and real-time wheat crop stage recognition. The data collection utilized a diverse image dataset covering seven growth phases 'Crown Root', 'Tillering', 'Mid Vegetative', 'Booting', 'Heading', 'Anthesis', and 'Milking', comprising 4496 images. The collected image dataset underwent rigorous preprocessing and advanced data augmentation to refine and minimize biases. This study employed deep and transfer learning models, including MobileNetV2, DenseNet-121, NASNet-Large, InceptionV3, and a convolutional neural network (CNN) for performance comparison. Experimental evaluations demonstrated that the transfer model MobileNetV2 achieved 95% accuracy, DenseNet-121 achieved 94% accuracy, NASNet-Large achieved 76% accuracy, InceptionV3 achieved 74% accuracy, and the CNN achieved 68% accuracy. The proposed novel hybrid approach, MobDenNet, that synergistically merges the architectures of MobileNetV2 and DenseNet-121 neural networks, yields highly accurate results with precision, recall, and an F1 score of 99%. We validated the robustness of the proposed approach using the k-fold cross-validation. The proposed research ensures the detection of growth stages with great promise for boosting agricultural productivity and management practices, empowering farmers to optimize resource distribution and make informed decisions.

关键词： Precision agriculture Agricultural system Wheat growth prediction Hybrid neural network image processing deep learning

来源：评论

学校读者我要写书评

暂无评论

Leveraging Augmented reality,Semantic-Segmentation,and VANETs for Enhanced Driver’s Safety Assistance

引用

Computers, Materials & Continua 2024年第1期78卷 1443-1460页

作者： Sitara Afzal Imran Ullah Khan Irfan Mehmood Jong Weon Lee Mixed Reality and Interaction Lab Department of SoftwareSejong UniversitySeoul05006Korea Faculty of Engineering and Digital Technologies School of Computer ScienceAI and ElectronicsUniversity of BradfordBradfordUK

Overtaking is a crucial maneuver in road transportation that requires a clear view of the road ***,limited visibility of ahead vehicles can often make it challenging for drivers to assess the safety of overtaking maneuvers,leading to accidents and *** this paper,we consider atrous convolution,a powerful tool for explicitly adjusting the field-of-view of a filter as well as controlling the resolution of feature responses generated by deep Convolutional Neural Networks in the context of semantic image *** article explores the potential of seeing-through vehicles as a solution to enhance overtaking ***-through vehicles leverage advanced technologies such as cameras,sensors,and displays to provide drivers with a real-time view of the vehicle ahead,including the areas hidden from their direct line of *** address the problems of safe passing and occlusion by huge vehicles,we designed a see-through vehicle system in this study,we employed a windshield display in the back car together with cameras in both *** server within the back car was used to segment the car,and the segmented portion of the car displayed the video from the front *** see-through system improves the driver’s field of vision and helps him change lanes,cross a large car that is blocking their view,and safely overtake other *** network was trained and tested on the Cityscape dataset using semantic *** transparent technique will instruct the driver on the concealed traffic situation that the front vehicle has *** our findings,we have achieved 97.1% *** article also discusses the challenges and opportunities of implementing see-through vehicles in real-world scenarios,including technical,regulatory,and user acceptance factors.

关键词： Overtaking safety augmented reality VANET V2V deep learning

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：