Article
Computer Science, Information Systems
Zhongyong Wang, Dongzhe Sun, Kexian Gong, Wei Wang, Peng Sun
Summary: This paper proposes a lightweight convolutional neural network for automatic modulation classification task, with a focus on reducing model complexity by designing depthwise separable convolution residual architecture and using global depthwise convolution for feature reconstruction. Experimental results show significant savings in model parameters and inference time compared to recent works.
Article
Remote Sensing
Ke Zhang, Inuwa Mamuda Bello, Yu Su, Jingyu Wang, Ibrahim Maryam
Summary: This article proposes a lightweight multiscale segmentation framework that achieves high accuracy pixel prediction with relatively low computational overhead by embedding sparse network architecture and depthwise separable convolution at the multiscale level.
INTERNATIONAL JOURNAL OF REMOTE SENSING
(2022)
Article
Computer Science, Artificial Intelligence
Libo Chang, Shengbing Zhang, Huimin Du, Zhonglun You, Shiyu Wang
Summary: This paper proposes a lightweight network structure based on depthwise separable convolution, incorporating specific convolution modules to improve accuracy in object detection by utilizing specific position information. Through experiments, the method shows effective model size compression and reduced computational latency on COCO dataset, while achieving better accuracy on small object detection on PASCAL VOC2007 dataset. The proposed approach outperforms lightweight algorithms based on knowledge distillation or depthwise separable convolution in terms of accuracy, parameter count, and real-time performance.
JOURNAL OF REAL-TIME IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Jun-Gi Jang, Chun Quan, Hyun Dong Lee, U. Kang
Summary: This paper proposes Falcon, an accurate and lightweight method to compress CNN based on depthwise separable convolution. Falcon interprets existing convolution methods based on depthwise separable convolution using a proposed mathematical formulation called generalized elementwise product (GEP). Experimental results show that Falcon achieves higher accuracy than existing methods while reducing the number of parameters and FLOPs of standard convolution.
KNOWLEDGE AND INFORMATION SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Jiaxing He, Xiaodan Wang, Yafei Song, Qian Xiang
Summary: To solve the issue of feature extraction in network intrusion detection caused by large-scale high-dimensional traffic data, we propose PyDSC-IDS, a method based on the variational Gaussian model and one-dimensional Pyramid Depthwise Separable Convolution neural network. PyDSC-IDS uses VGM and OneHot encode technologies to preprocess the original dataset and decompose complex features into simpler ones. The experimental results demonstrate the effectiveness of PyDSC-IDS in improving detection accuracy and reducing network complexity.
Article
Computer Science, Artificial Intelligence
Shen Yan, Haidong Shao, Jie Wang, Xinyu Zheng, Bin Liu
Summary: This paper proposes a lightweight fault diagnosis framework called LiConvFormer, which uses a separable multi scale convolution block and a broadcast self-attention block to extract local and global features of signals respectively, addressing the issues of complex collaborative models and limited applications.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)
Article
Environmental Sciences
Yu Sun, Liang Huang, Junsan Zhao, Xiaoxiang Li, Mulan Qiu
Summary: This study proposes a deep learning-based method for bridge detection, which utilizes a depth-wise separable multiscale feature fusion network. The experimental results demonstrate that the proposed method achieves high accuracy and speed, making it suitable for high-precision and fast bridge detection tasks.
GEOCARTO INTERNATIONAL
(2022)
Article
Engineering, Electrical & Electronic
Chenghong Xiao, Shuyuan Yang, Zhixi Feng
Summary: In this article, a novel end-to-end automatic modulation classification (AMC) model called complex-valued depthwise separable convolutional neural network (CDSCNN) is proposed, which adopts complex-valued operation units for tailored feature learning for AMC. With an overall accuracy of 62.63% on the RadioML2016.10a dataset, CDSCNN outperforms its counterparts by 1%-11%. After fine-tuning on the RadioML2016.10b dataset, the overall accuracy reaches 63.15%, demonstrating the robust recognition and generalization capability of CDSCNN. Moreover, CDSCNN exhibits lower model complexity compared to other methods.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Computer Science, Interdisciplinary Applications
Junxiu Liu, Mingxing Li, Yuling Luo, Su Yang, Wei Li, Yifei Bi
Summary: Neuroimaging methods are employed for diagnosing Alzheimer's disease, with recent research focusing on machine learning algorithms inspired by computer vision with deep learning. However, limitations such as the need for a large number of training images and powerful computers hinder widespread usage of AD diagnosis based on machine learning. Proposed deep separable convolutional neural network model improves efficiency by greatly reducing parameters and computing costs compared to traditional neural networks.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2021)
Article
Chemistry, Multidisciplinary
Sheng Yuan, Zhao Qiu, Peipei Li, Yuqi Hong
Summary: This paper proposes a new breast tumor segmentation network, RMAU-Net, which addresses the problems of existing networks by combining residual depthwise separable convolution and multi-scale channel attention gate, and prevents information loss by using Patch Merging operation. Experiments demonstrate that the method has superior segmentation performance and better generalization.
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Artificial Intelligence
Jingdong Yang, Lei Zhang, Xinjun Tang, Man Han
Summary: A novel lightweight CNN model, CodnNet, is proposed in this paper for quick detection of COVID-19 infection. The model improves feature reuse and receptive field range through effective dense connections and depthwise separable convolution, resulting in better classification accuracy and generalization performance.
APPLIED SOFT COMPUTING
(2022)
Article
Mathematical & Computational Biology
Sohaib Asif, Ming Zhao, Xuehan Chen, Yusen Zhu
Summary: Kidney stone disease is a common and serious health problem worldwide. This study proposes a lightweight and high-performance model, StoneNet, for the detection of kidney stones. Experimental results show that StoneNet outperforms other models in terms of accuracy and complexity, and can assist radiologists in faster diagnosis of kidney stones.
INTERDISCIPLINARY SCIENCES-COMPUTATIONAL LIFE SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Xiaoyu Kong, Ke Zhang
Summary: Human behavior is influenced by emotions, and predicting behavior through emotion classification from text is significant for decision-making. Efficiently extracting emotional tendencies from text data is a challenge, but a upgraded CNN model proposed in this study improves the downsides and shows better performance in sentiment analysis tasks.
PEERJ COMPUTER SCIENCE
(2023)
Article
Multidisciplinary Sciences
Shuo Pan, Hai Yan, Zhuo Liu, Ning Chen, Yinghao Miao, Yue Hou
Summary: Texture is a crucial characteristic of roads, and recognizing pavement texture is important for road maintenance professionals. This paper proposes a few-shot learning model based on the Siamese network for pavement texture recognition. The model achieved high accuracy in classifying different pavement textures and created lightweight models that save storage and training time.
PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY A-MATHEMATICAL PHYSICAL AND ENGINEERING SCIENCES
(2023)
Article
Computer Science, Information Systems
Lu Yang, Qing Song, Zimeng Fan, Chun Liu, Mengjie Hu
Summary: Activation function plays a crucial role in neural networks, as it can enhance accuracy and convergence speed. This paper investigates the problem of information loss caused by activation functions in lightweight networks and proposes a solution using activation functions with negative values. The authors present a method to minimize changes to existing networks by replacing ReLU with Swish at appropriate positions in lightweight networks, termed as enriching activation. They also introduce a new activation function called (H)-SwishX for enriching activation, which learns significant maximal values in each layer of the network to reduce accuracy reduction during lightweight network quantization. The proposed enriching activation scheme demonstrates performance improvements on popular lightweight networks, verified using CIFAR-10 and ImageNet datasets, and shows potential for transfer learning as well, evaluated using MSCOCO object detection.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Engineering, Civil
Guang Yuan, Yanyan Chen, Lishan Sun, Jianhui Lai, Tongfei Li, Zhuo Liu
JOURNAL OF ADVANCED TRANSPORTATION
(2020)
Article
Mathematics, Interdisciplinary Applications
Chengcheng Song, Yanyan Chen, Ning Chen, Zhuo Liu, Xuzhen Zhu, Wei Wang
Article
Construction & Building Technology
Jianlin Jia, Yanyan Chen, Ning Chen, Hui Yao, Yongxing Li, Zhuo Liu
ADVANCES IN CIVIL ENGINEERING
(2020)
Article
Engineering, Civil
Ning Chen, Zijin Xu, Zhuo Liu, Yihan Chen, Yinghao Miao, Qiuhan Li, Yue Hou, Linbing Wang
Summary: The adaption to roads with different macro-textures is of great significance for autonomous driving technologies. In this study, pavement texture recognition using deep learning approaches was explored, and the data augmentation was performed using a Generative Adversarial Network (GAN) model. The results showed that deep learning methods outperformed manual classification and traditional machine learning methods in pavement texture recognition.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Engineering, Civil
Yue Hou, Hongyu Shi, Ning Chen, Zhuo Liu, Han Wei, Qiang Han
Summary: This paper presents an engineering approach that integrates transfer learning with lightweight models for classifying and detecting concrete bridge distress. Experimental results show that this approach has the potential for application in intelligent transportation infrastructure maintenance.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Engineering, Civil
Yue Hou, Shuo Liu, Dandan Cao, Bo Peng, Zhuo Liu, Wenjuan Sun, Ning Chen
Summary: This study proposes an intelligent method based on limited field images to classify pavement crack images. By using data augmentation and crack extraction as preprocessing steps, combined with model selection and hyper-parameter tuning, the final approach significantly improves the test accuracy.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Transportation
Jiaomin Wei, Yang Wang, Zhuo Liu, Yanyan Chen
Summary: To analyze the correlation between the built environment and dockless bike-sharing trips connecting to urban metro stations, a framework augmented by big data is proposed. The results show that the presence of greenery and barriers has a significant positive influence on trip density and cycling speed, while the presence of streetlights and signals has a negative impact on trip density and cycling speed.
JOURNAL OF TRANSPORT AND LAND USE
(2023)
Article
Multidisciplinary Sciences
Shuo Pan, Hai Yan, Zhuo Liu, Ning Chen, Yinghao Miao, Yue Hou
Summary: Texture is a crucial characteristic of roads, and recognizing pavement texture is important for road maintenance professionals. This paper proposes a few-shot learning model based on the Siamese network for pavement texture recognition. The model achieved high accuracy in classifying different pavement textures and created lightweight models that save storage and training time.
PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY A-MATHEMATICAL PHYSICAL AND ENGINEERING SCIENCES
(2023)
Article
Engineering, Civil
Zhuo Liu, Sichun Li, Xiaoxiong Zhao, Zhenbao Wang, Yanyan Chen
Summary: This study proposes a novel framework to analyze the accessibility of urban medical departments for older adults, taking Beijing as a case study. The results show that medical accessibility is generally higher in downtown areas and decreases towards the periphery. Cardiovascular, respiratory, and orthopedic patients face greater difficulty in accessing appropriate medical resources. Significant inequities in medical accessibility are observed among subdistricts in different locations.
JOURNAL OF URBAN PLANNING AND DEVELOPMENT
(2023)
Article
Engineering, Civil
Zijin Xu, Xin Yu, Zhuo Liu, Song Zhang, Qinxia Sun, Ning Chen, Haotian Lv, Dawei Wang, Yue Hou
Summary: This study uses Ground Penetrating Radar (GPR) to monitor the safety of transportation infrastructure and proposes a deep data augmentation method to address the small sample size and unbalanced dataset issues. With the help of intelligent classification techniques, accurate identification of subgrade distresses is achieved.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Ergonomics
Shaohua Wang, Yanyan Chen, Jianling Huang, Zhuo Liu, Jia Li, Jianming Ma
JOURNAL OF SAFETY RESEARCH
(2020)
Article
Computer Science, Information Systems
Tongfei Li, Yanyan Chen, Zhen Wang, Zhuo Liu, Rui Ding, Shuqi Xue
Article
Construction & Building Technology
Jia Liang, Qipeng Zhang, Xingyu Gu
Summary: A lightweight PCSNet-based segmentation model is developed to address the issues of insufficient performance in feature extraction and boundary loss information. The introduction of generalized Dice loss improves prediction performance, and the visualization of class activation mapping enhances model interpretability.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Construction & Building Technology
Gilsu Jeong, Minhyuk Jung, Seongeun Park, Moonseo Park, Changbum Ryan Ahn
Summary: This study introduces a contextual audio-visual approach to recognize multi-equipment activities in tunnel construction sites, improving monitoring effectiveness. Tested against real-world operation data, the model achieved remarkable results, emphasizing the potential of contextual multimodal models in enhancing operational efficiency in complex construction sites.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Construction & Building Technology
Jin Wang, Zhigao Zeng, Pradip Kumar Sharma, Osama Alfarraj, Amr Tolba, Jianming Zhang, Lei Wang
Summary: This study presents a dual-path network for pavement crack segmentation, combining Convolutional Neural Network (CNN) and transformer. A lightweight CNN encoder is used for local feature extraction, while a novel transformer encoder integrates high-low frequency attention mechanism and efficient feedforward network for global feature extraction. Additionally, a complementary fusion module is introduced to aggregate intermediate features extracted from both encoders. Evaluation on three datasets confirms the superior performance of the proposed network.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Construction & Building Technology
Pierre Gilibert, Romain Mesnil, Olivier Baverel
Summary: This paper introduces a flexible method for crafting 2D assemblies adaptable to various geometric assumptions in the realm of sustainable construction. By utilizing digital fabrication technologies and optimization approaches, precise control over demountable buildings can be achieved, improving mechanical performance and sustainability.
AUTOMATION IN CONSTRUCTION
(2024)
Review
Construction & Building Technology
Jorge Loy-Benitez, Myung Kyu Song, Yo-Hyun Choi, Je-Kyum Lee, Sean Seungwon Lee
Summary: This paper discusses the advancement of tunnel boring machines (TBM) through the application of artificial intelligence. It highlights the significance of AI-based management subsystems for automatic TBM operations and presents recent contributions in this field. The paper evaluates modeling, monitoring, and control subsystems and suggests research paths for integrating existing management subsystems into TBM automation.
AUTOMATION IN CONSTRUCTION
(2024)
Review
Construction & Building Technology
Alireza Shamshiri, Kyeong Rok Ryu, June Young Park
Summary: This paper reviews the application of text mining and natural language processing in the construction field, highlighting the need for automation and minimizing manual tasks. The study identifies potential research opportunities in strengthening overlooked construction aspects, coupling diverse data formats, and leveraging pre-trained language models and reinforcement learning.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Construction & Building Technology
Zhengyi Chen, Hao Wang, Keyu Chen, Changhao Song, Xiao Zhang, Boyu Wang, Jack C. P. Cheng
Summary: This study proposes an improved coverage path planning system that leverages building information modeling and robotic configurations to optimize coverage performance in indoor environments. Experimental validation shows the effectiveness and applicability of the system. Future research will focus on further enhancing coverage ratio and optimizing computation time.
AUTOMATION IN CONSTRUCTION
(2024)
Review
Construction & Building Technology
Yonglin Fu, Junjie Chen, Weisheng Lu
Summary: This study presents a review of human-robot collaboration (HRC) in modular construction manufacturing (MCM), focusing on tasks, human roles, and interaction levels. The review found that HRC solutions are applicable to various MCM tasks, with a primary focus on timber component production. It also revealed the diverse collaborative roles humans can play and the varying levels of interaction with robots.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Construction & Building Technology
Qiong Liu, Shengbo Cheng, Chang Sun, Kailun Chen, Wengui Li, Vivian W. Y. Tam
Summary: This paper presents an approach to enhance the path-following capability of concrete printing by integrating steel cables into the printed mortar strips, and validates the feasibility and effectiveness of this approach through experiments.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Construction & Building Technology
Honghu Chu, Lu Deng, Huaqing Yuan, Lizhi Long, Jingjing Guo
Summary: The study proposes a method called Cascade CATransUNet for high-resolution crack image segmentation. This method combines the coordinate attention mechanism and self-cascaded design to accurately segment cracks. Through a customized feature extraction architecture and an optimized boundary loss function, the proposed method achieves impressive segmentation performance on HR images and demonstrates its practicality in UAV crack detection tasks.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Construction & Building Technology
Daniel Lamas, Andres Justo, Mario Soilan, Belen Riveiro
Summary: This paper introduces a new method for creating synthetic point clouds of truss bridges and demonstrates the effectiveness of a deep learning approach for semantic and instance segmentation of these point clouds. The proposed methodology has significant implications for the development of automated inspection and monitoring systems for truss bridges.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Construction & Building Technology
Kahyun Jeon, Ghang Lee, Seongmin Yang, Yonghan Kim, Seungah Suh
Summary: This study proposes two enhanced unsupervised text classification methods for domain-specific non-English text. The results of the tests show that these methods achieve excellent performance on Korean building defect complaints, outperforming state-of-the-art zero-shot and few-shot text classification methods, with minimal data preparation effort and computing resources.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Construction & Building Technology
Yoonhwa Jung, Julia Hockenmaier, Mani Golparvar-Fard
Summary: This study introduces a transformer-based natural language processing model, UNIfORMATBRIDGE, that automatically labels activities in a project schedule with Uniformat classification. Experimental results show that the model performs well in matching unstructured schedule data to Uniformat classifications. Additionally, the study highlights the importance of this method in developing new techniques.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Construction & Building Technology
De-Graft Joe Opoku, Srinath Perera, Robert Osei-Kyei, Maria Rashidi, Keivan Bamdad, Tosin Famakinwa
Summary: This paper introduces a digital twin technology combining Building Information Modelling and the Internet of Things for the construction industry, aiming to optimize building conditions. The technology is implemented in a university library, successfully achieving real-time data capture and visual representation of internal conditions.
AUTOMATION IN CONSTRUCTION
(2024)
Article
Construction & Building Technology
Zaolin Pan, Yantao Yu
Summary: The construction industry faces safety and workforce shortages globally, and worker-robot collaboration is seen as a solution. However, robots face challenges in recognizing worker intentions in construction. This study tackles these challenges by proposing a fusion method and investigating the best granularity for recognizing worker intentions. The results show that the proposed method can recognize multi-granular worker intentions effectively, contributing to seamless worker-robot collaboration in construction.
AUTOMATION IN CONSTRUCTION
(2024)