Article
Computer Science, Artificial Intelligence
Zheng Zou, Peng Zhao, Xuefeng Zhao
Summary: This paper presents a virtual restoration method for weathered beams in the Forbidden City using multiple deep learning algorithms to repair colored paintings. By dividing the paintings into three parts for restoration, it provides reference and guidance for traditional manual restoration, reducing complexity and repetitive work.
ADVANCED ENGINEERING INFORMATICS
(2021)
Article
Computer Science, Interdisciplinary Applications
Atiqur Rahman, Zheng Yi Wu, Rony Kalfarisi
Summary: This paper introduces a semantic segmentation deep learning approach combined with an efficient image labeling tool for rapidly preparing large training data sets, and effectively detecting, segmenting, and evaluating corrosion in images.
JOURNAL OF COMPUTING IN CIVIL ENGINEERING
(2021)
Review
Environmental Sciences
Su Yang, Miaole Hou, Songnian Li
Summary: This paper provides a comprehensive literature review on the acquisition of point cloud data and semantic segmentation algorithms in the cultural heritage field. It discusses the current trends, advantages, disadvantages, and applications of various algorithms, including region growing, model fitting, unsupervised clustering, supervised machine learning, and deep learning. Additionally, it summarizes the public benchmark datasets related to cultural heritage and presents the problems and future development trends of 3D point cloud semantic segmentation in this field.
Article
Ecology
Serkan Kartal
Summary: The study compares deep learning models for segmenting clover, grass, and weeds in RGB images, with the FPNInceptionresnetv2 model achieving the highest segmentation accuracy of 76.7%. This suggests great potential for the use of deep convolutional neural networks in plant species segmentation from RGB images.
ECOLOGICAL INFORMATICS
(2021)
Article
Mathematics, Interdisciplinary Applications
Abdulmalik Aldawsari, Syed Adnan Yusuf, Riad Souissi, Muhammad AL-Qurishi
Summary: Automated assessment of car damage is a challenge in the auto repair and damage assessment industries. This study explores instance segmentation methods to find the best-performing models for identifying car parts. The YOLACT-based method showed better performance in part localization and segmentation, while SipMask++ had better accuracy in object detection for a workshop repair dataset.
Article
Construction & Building Technology
Iason Katsamenis, Nikolaos Doulamis, Anastasios Doulamis, Eftychios Protopapadakis, Athanasios Voulodimos
Summary: This paper presents an architecture for simultaneous precise localization and classification of corrosion and rust grade recognition. By using a lightweight deep learning model and a novel data projection scheme, the proposed method achieves excellent performance in corrosion detection, localization, and classification, with shorter computation time compared to other methods.
AUTOMATION IN CONSTRUCTION
(2022)
Article
Computer Science, Artificial Intelligence
Tianfei Zhou, Fatih Porikli, David J. Crandall, Luc Van Gool, Wenguan Wang
Summary: Video segmentation is crucial in various practical applications such as enhancing visual effects in movies, understanding scenes in autonomous driving, and creating virtual background in video conferencing. Deep learning-based approaches have shown promising performance in video segmentation. This survey comprehensively reviews two main research lines - generic object segmentation and video semantic segmentation - by introducing their task settings, background concepts, need, development history, and challenges. Representative literature and datasets are also discussed, and the reviewed methods are benchmarked on well-known datasets. Open issues and opportunities for further research are identified, and a public website is provided to track developments in this field: https://github.com/tfzhou/VS-Survey.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Zihao Wu, Yunchao Tang, Bo Hong, Bingqiang Liang, Yuping Liu
Summary: This study proposes a lightweight image segmentation network for dam crack detection. By improving the network structure and parameters, it achieves higher accuracy in crack segmentation. Experimental results demonstrate the potential of this method for crack detection and its wide application in engineering practice.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2023)
Review
Computer Science, Artificial Intelligence
Saeid Asgari Taghanaki, Kumar Abhishek, Joseph Paul Cohen, Julien Cohen-Adad, Ghassan Hamarneh
Summary: This review categorizes leading deep learning-based medical and non-medical image segmentation solutions into six main groups and provides a comprehensive review of each group's contributions. It analyzes the limitations of current approaches and presents potential future research directions for improving semantic image segmentation.
ARTIFICIAL INTELLIGENCE REVIEW
(2021)
Review
Computer Science, Artificial Intelligence
Jigang Tang, Songbin Li, Peng Liu
Summary: This paper provides a comprehensive review of lane detection methods, covering background introduction, method classification, network architectures, loss functions, and method comparisons. Current challenges and future research directions are also discussed.
PATTERN RECOGNITION
(2021)
Article
Agronomy
Yun Peng, Aichen Wang, Jizhan Liu, Muhammad Faheem
Summary: This article studied the segmentation of grape clusters with different varieties using three state-of-the-art semantic segmentation networks. The results showed that DeepLabv3+ combined with transfer learning performed the best, while L*a*b representation achieved the highest IoU. Image enhancement and distance between clusters and camera also had significant impacts on segmentation performance, with closer distances leading to better results.
Article
Agriculture, Multidisciplinary
Hanwen Kang, Xing Wang
Summary: This paper proposes a deep-learning-based segmentation method to perform accurate semantic segmentation on fused data from a LiDAR-Camera visual sensor. It solves the problems of multi-sensor data fusion and network training under imbalanced class conditions. The experiment results show that the proposed method can perform accurate segmentation in real orchard environments.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2023)
Article
Computer Science, Artificial Intelligence
Zhixin Guo, Wenzhi Liao, Yifan Xiao, Peter Veelaert, Wilfried Philips
Summary: This study addresses the issue of time-consuming pixel-level semantic segmentation annotations in pedestrian detection by using weak segmentation masks automatically generated from depth images. It improves the performance of pedestrian detectors and further enhances detection performance. The effectiveness of the proposed method is demonstrated through extensive experiments, producing good-quality pedestrian segmentation results without the use of pixel-level segmentation annotations during training.
PATTERN RECOGNITION
(2021)
Article
Engineering, Electrical & Electronic
Xinye Li, Ding Chen
Summary: This article summarizes the basic ideas of panoptic segmentation method based on deep learning and classifies the current image panoptic segmentation into four categories, analyzing the characteristics and limitations of each method and comparing segmentation effects. It also involves video panoptic segmentation and LiDAR data panoptic segmentation, and future research directions are prospected.
DIGITAL SIGNAL PROCESSING
(2022)
Article
Computer Science, Artificial Intelligence
Zizi Chen, Gary W. Small
Summary: The United States Environmental Protection Agency's program provides infrared remote sensing capabilities to assist first responders in managing chemical releases. They used a multi-imagining system and a deep learning model to automate the detection of chemical plumes, achieving high accuracy and false detection rates below 0.02%.
NEURAL COMPUTING & APPLICATIONS
(2022)
Review
Computer Science, Artificial Intelligence
Chnoor M. Rahman, Tarik A. Rashid, Abeer Alsadoon, Nebojsa Bacanin, Polla Fattah, Seyedali Mirjalili
Summary: This paper provides a comprehensive investigation of the dragonfly algorithm in the engineering field. It discusses the overview and modifications of the algorithm, surveys its applications in engineering, and compares its performance with other algorithms. The results show that the dragonfly algorithm performs excellently in small to intermediate applications. The purpose of this research is to assist other researchers in studying and utilizing the algorithm to optimize engineering problems.
EVOLUTIONARY INTELLIGENCE
(2023)
Review
Energy & Fuels
Nebojsa Bacanin, Catalin Stoean, Miodrag Zivkovic, Miomir Rakic, Roma Strulak-Wojcikiewicz, Ruxandra Stoean
Summary: An effective energy oversight is a global concern, especially with recent increasing stringency. Machine learning and deep learning approaches have shown high accuracy in energy load and consumption prediction, but few recent methods focus on parameter tuning for better results. This study develops and tunes a long short-term memory (LSTM) DL model for multivariate time-series forecasting of electricity load, using a benchmark dataset from Europe. The results serve as a benchmark for hybrid LSTM-optimization methods in energy time-series forecasting. The study highlights the importance of parameter tuning for improved results using metaheuristics, with the worst-performing metaheuristic still outperforming grid search.
Article
Computer Science, Information Systems
K. Venkatachalam, Zaoli Yang, Pavel Trojovsky, Nebojsa Bacanin, Muhammet Deveci, Weiping Ding
Summary: Human activity recognition (HAR) is an emerging field that identifies human actions in different settings. This study proposes a hybrid model combining one-dimensional convolutional neural network and long short term memory (LSTM) classifier to improve the performance of HAR. The UCI-HAR dataset is used for experimental research.
INFORMATION SCIENCES
(2023)
Article
Engineering, Electrical & Electronic
Marko Sarac, Milos Mravik, Dijana Jovanovic, Ivana Strumberger, Miodrag Zivkovic, Nebojsa Bacanin
Summary: COVID-19 is a respiratory system disorder that has caused pneumonia outbreaks globally. Computed tomography (CT) has played a crucial role in diagnosing the disease. This study proposes a deep learning model called DEPNet to predict COVID-19 cases based on CT images.
JOURNAL OF ELECTRONIC IMAGING
(2023)
Article
Computer Science, Information Systems
Milos Mravik, Marko Sarac, Nebojsa Bacanin, Sasa Adamovic
Summary: This paper examines the impact of new approaches to distance learning during the Covid-19 pandemic, with a focus on comparing the quality of online teaching to face-to-face teaching. The presented results are based on empirical research conducted over a period of 2 years with a large group of students. The study finds that both professors and students encountered self-imposed obstacles, as well as pedagogical, technical, and financial or organizational barriers. The obtained results are further supported by conducting relevant hypothesis tests.
JOURNAL OF INTERNET TECHNOLOGY
(2023)
Article
Multidisciplinary Sciences
M. G. Dinesh, Nebojsa Bacanin, S. S. Askar, Mohamed Abouhawwash
Summary: Pancreatic cancer is often diagnosed at an advanced stage, leading to high mortality rates. Therefore, automated systems that can detect cancer early are crucial. This research aims to predict pancreatic cancer early using deep learning and metaheuristic techniques, analyzing medical imaging data and identifying vital features and cancerous growths.
SCIENTIFIC REPORTS
(2023)
Article
Multidisciplinary Sciences
Nebojsa Bacanin, Nebojsa Budimirovic, K. Venkatachalam, Hothefa Shaker Jassim, Miodrag Zivkovic, S. S. Askar, Mohamed Abouhawwash
Summary: With the rapid growth of stored data in datasets, extracting crucial information becomes difficult. This research presents a novel quasi-reflection learning algorithm - firefly search, an enhanced version of the original arithmetic optimization algorithm. The proposed algorithm is tested on benchmark functions, standard datasets, and a Corona disease dataset, and the experimental results verify its improvements and statistical significance.
Article
Computer Science, Software Engineering
Luka Jovanovic, Dijana Jovanovic, Milos Antonijevic, Bosko Nikolic, Nebojsa Bacanin, Miodrag Zivkovic, Ivana Strumberger
Summary: This research proposes a hybrid approach based on an improved metaheuristics algorithm to optimize the XGBoost machine learning model for enhancing Web security. Evaluations on three publicly available phishing website datasets show that the proposed solution outperforms other methods and represents a perspective solution in the domain of web security.
JOURNAL OF WEB ENGINEERING
(2023)
Article
Environmental Sciences
Charli Sitinjak, Vladimir Simic, Rozmi Ismail, Nebojsa Bacanin, Charles Musselwhite
Summary: Effective end-of-life vehicle (ELV) management is crucial for minimizing the environmental and health impacts of Indonesia's growing automotive industry. However, proper ELV management has received limited attention. Our qualitative study identified barriers to effective ELV management in Indonesia's automotive sector, including inadequate regulation and enforcement, insufficient infrastructure and technology, low education and awareness, and a lack of financial incentives. We recommend a comprehensive and integrated approach involving coordination among government, industry, and stakeholders to address these barriers and develop sustainable ELV management policies and decisions.
ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
(2023)
Article
Computer Science, Information Systems
B. Saravana Balaji, Wieslaw Paja, Milos Antonijevic, Catalin Stoean, Nebojsa Bacanin, Miodrag Zivkovic
Summary: Smart cities consist of intelligent industrial devices that improve people's lives and save lives. Intelligent remote patient monitoring predicts the patient's condition. Internet of Things (IoT), artificial intelligence (AI), and cloud computing have enhanced the healthcare industry. Edge computing accelerates patient data transmission and ensures latency, reliability, and response time. However, the transmission of large amounts of patient data may lead to IoT data security vulnerabilities, posing concerns and challenges. This research proposes a secure, scalable, and responsive patient monitoring system. The model uses lightweight attribute-based encryption (LABE) to encrypt and decrypt IoT patient data for cloud-based protection. Edge servers are situated between the IoT and cloud for improved quality of service (QoS) and patient diagnosis. The deep belief network (DBN) predicts and monitors patient health, while the bat optimization algorithm (BOA) optimizes hyperparameters. Deep belief is used to identify hyperparameters, and BOA is applied for optimization. Swarm intelligence enhances prediction results and edge-cloud reaction time. A simulated environment evaluates the secure patient health monitoring system for efficiency, security, and efficacy. The proposed model offers effective patient remote health monitoring through a secure edge-cloud-IoT environment, with improved accuracy (97.9%), precision (95.6%), recall (94.6%), F1-score (94.9%), and false discovery rate (0.06%).
HUMAN-CENTRIC COMPUTING AND INFORMATION SCIENCES
(2023)
Article
Computer Science, Information Systems
Sara Boskovic, Libor Svadlenka, Stefan Jovcic, Momcilo Dobrodolac, Vladimir Simic, Nebojsa Bacanin
Summary: This paper introduces a new subjective technique called FullEX for evaluating the importance of criteria in LMD courier selection. Through evaluation, on-time delivery is considered the most important criterion for sustainable LMD courier selection.
Article
Chemistry, Multidisciplinary
Mohamed Salb, Luka Jovanovic, Nebojsa Bacanin, Milos Antonijevic, Miodrag Zivkovic, Nebojsa Budimirovic, Laith Abualigah
Summary: This paper addresses the critical security challenges in the internet of things (IoT) landscape by proposing an innovative solution that combines convolutional neural networks (CNNs) and the XGBoost model for intrusion detection. By customizing the reptile search algorithm for hyperparameter optimization, the methodology provides a resilient defense against emerging threats in IoT security. The introduced algorithm constructed models with the best performance in both experiments, and its outcomes have been statistically evaluated and analyzed for feature importance using Shapley additive explanations.
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Artificial Intelligence
Mihailo Todorovic, Nemanja Stanisic, Miodrag Zivkovic, Nebojsa Bacanin, Vladimir Simic, Erfan Babaee Tirkolaee
Summary: This study aims to create a machine learning model that can predict opinions in external audits and surpass the benchmark set in prior research. The study compares the performance of different algorithms in two scenarios and finds improvement through optimized hyperparameter tuning and the use of an iterative algorithm.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Information Systems
Sara Boskovic, Libor Svadlenka, Stefan Jovcic, Momcilo Dobrodolac, Vladimir Simic, Nebojsa Bacanin
Summary: This article introduces a new multi-criteria decision-making method, called the AROMAN method, which combines linear and vector normalization techniques to obtain accurate data structures and develops an original final ranking equation. Comparative analysis shows a high level of confidence and stability in the AROMAN method in the decision-making field.
Article
Archaeology
Justin J. L. Kimball, Ruben With, Christian Lochsen Rodsrud
Summary: Micro-CT (μCT) has been increasingly used in the cultural heritage sector to understand past cultures and their materials. In the case of the Gjellestad ship from the Viking Age, μCT was used to document and conserve the deteriorated organic and metallic materials. A georeferencing system was developed to retain important stratigraphic and position information, allowing for spatial positioning using 3D GIS. The use of μCT has shown positive impact on the documentation, conservation, and reconstruction of cultural heritage.
JOURNAL OF CULTURAL HERITAGE
(2024)
Review
Archaeology
S. Sylaiou, P. Dafiotis, D. Koukopoulos, K. Koukoulis, R. Vital, A. Antoniou, Chr. Fidas
Summary: This manuscript discusses the growing importance of Extended Reality (XR) in art exhibitions. It explores the technologies used, design issues, evaluation metrics, and aims of XR exhibitions. The research focuses on the current types of technologies used, primary design considerations, and methods to enhance user experience. The paper also examines evaluation criteria and the use of virtual humans for increased engagement. Additionally, it thoroughly discusses parameters affecting user experience and offers suggestions for optimizing design and future directions.
JOURNAL OF CULTURAL HERITAGE
(2024)
Article
Archaeology
Carlo Battini, Umberto Ferretti, Giorgia De Angelis, Roberto Pierdicca, Marina Paolanti, Ramona Quattrini
Summary: This research presents a method for recognizing historical building elements using a deep learning system. By leveraging synthetic point clouds to generate 3D models, the proposed approach achieves high accuracy in the experiments conducted on a newly synthetic dataset.
JOURNAL OF CULTURAL HERITAGE
(2024)
Article
Archaeology
Jingwen Zhang, Tianlin Ren
Summary: In the era of digital information, metadata and ontology technology have promoted the management and utilization of museum collections. This study focused on ancient Chinese ceramics and proposed solutions for the shortcomings in their information management through metadata analysis and ontology construction. The structure of the ontology was visualized to provide a clearer understanding of ancient Chinese ceramics.
JOURNAL OF CULTURAL HERITAGE
(2024)
