Article
Computer Science, Information Systems
Hassan Imani, Md Baharul Islam, Masum Shah Junayed, Tarkan Aydin, Nafiz Arica
Summary: Recently, 3D Convolutional Neural Networks (3D CNNs) have shown superior performance over 2D CNNs in video processing applications. In the field of Stereoscopic Video Quality Assessment (SVQA), 3D CNNs are used to extract spatio-temporal features from stereoscopic videos. Pre-trained 3D Residual Networks (3D ResNets) on the Kinetics dataset are fine-tuned to measure the quality of stereoscopic videos and propose a no-reference SVQA method. Experimental results on publicly available SVQA datasets demonstrate the effectiveness of the proposed transfer learning-based method.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Theory & Methods
Stefano Marrone, Cristina Papa, Carlo Sansone
Summary: This study investigates reducing the size of CNN by removing some neurons only from the fully connected layers before network training, and further compressing the network through weight quantization. The results show that it is possible to reduce the network size without statistically affecting performance.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2021)
Article
Computer Science, Information Systems
Jatin Karthik Tripathy, S. Sibi Chakkaravarthy, Suresh Chandra Satapathy, Madhulika Sahoo, V. Vaidehi
Summary: With the increasing use of online social media platforms, cyberbullying has become a concern. This paper focuses on textual comments and explores the challenges of contextual understanding in cyberbullying. The proposed ALBERT-based model achieved state-of-the-art results through fine-tuning, surpassing existing approaches.
MULTIMEDIA SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Sam Slade, Li Zhang, Haoqian Huang, Houshyar Asadi, Chee Peng Lim, Yonghong Yu, Dezong Zhao, Hanhe Lin, Rong Gao
Summary: This paper introduces a novel algorithm, neural inference search (NIS), for optimizing hyperparameters in deep learning segmentation models. NIS incorporates three new search behaviors, including maximized standard deviation velocity prediction, local best velocity prediction, and n-dimensional whirlpool search, to improve performance. Compared with state-of-the-art methods and other search algorithms, NIS-optimized models show significant improvements across multiple performance metrics on segmentation datasets.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Brandon Hobley, Michal Mackiewicz, Julie Bremner, Tony Dolphin, Riccardo Arosio
Summary: This article assessed the reliability of crowdsourced labels for estuarine vegetation and unvegetated sediment, and found that the accuracy of the labels was influenced by the expertise and familiarity of the participants. The results also confirmed that biases in participant annotation were propagated in the performance of the deep learning models. Additionally, it was shown that combining in situ and crowdsourced labels improved the performance of the models compared to using only in situ labels.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Lin Li, Xiaohui Lin, Tingting Liao, Rushan Ouyang, Meng Li, Jialin Yuan, Jie Ma
Summary: This study demonstrated the clinical application potential of a convolutional neural network (CNN)-based deep learning system in breast cancer screening and diagnosis in Asian women. The DL system showed higher sensitivity in mass detection compared to junior radiologists and was not affected by breast density.
QUANTITATIVE IMAGING IN MEDICINE AND SURGERY
(2023)
Review
Chemistry, Analytical
Jeon-Seong Kang, JinKyu Kang, Jung-Jun Kim, Kwang-Woo Jeon, Hyun-Joon Chung, Byung-Hoon Park
Summary: In recent years, deep learning has been extensively researched globally, particularly in training methods and network structures, and has proven highly effective in various tasks and applications. This paper summarizes the basic concepts of automated neural architecture search (NAS) and provides an overview of recent studies on its applications.
Article
Computer Science, Information Systems
Shivani Gaba, Ishan Budhiraja, Vimal Kumar, Sahil Garg, Georges Kaddoum, Mohammad Mehedi Hassan
Summary: This paper provides a detailed review of various deep learning architectures and models, with a focus on a specific convolutional neural network model. It discusses the working principles of convolutional neural networks and their components and presents various models from LeNet to AlexNet, GoogleNet, VGGNet, ResNet, DenseNet, Xception, PNAS/ENAS, and EfficientNet. The challenges associated with different network architectures are also summarized. The paper concludes with a discussion of the frameworks, datasets, applications, and accuracy of each model, serving as a future scope in the field.
COMPUTER COMMUNICATIONS
(2022)
Article
Computer Science, Information Systems
Samah A. Gamel, Esraa Hassan, Nora El-Rashidy, Fatma M. Talaat
Summary: The COVID-19 pandemic has had a significant impact on human migration worldwide, affecting transportation patterns in cities. This paper analyzes the relationship between COVID-19 and transportation using correlations and machine learning techniques, and introduces a Traffic Prediction Module (TPM) to predict the impact of COVID-19 on transportation. The results indicate a strong correlation between the spread of COVID-19 and transportation patterns, and the CNN has a high accuracy rate in predicting these impacts.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Geochemistry & Geophysics
Zhitao Zhao, Ping Tang, Lijun Zhao, Zheng Zhang
Summary: This study proposed a multiscale few-shot object detection approach for remote sensing images, which achieved better performance by optimizing the detector structure and training methodology.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Engineering, Electrical & Electronic
Xinglong Pei, Xiaoyang Zheng, Jinliang Wu
Summary: The paper introduces a novel Transformer convolution network (TCN) based on transfer learning, which has achieved highly accurate fault diagnosis. Experimental results demonstrate the robustness and effectiveness of the proposed method.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Computer Science, Information Systems
Lipeng Zhu, Junjie Lin
Summary: This article introduces a novel spatiotemporal correlation learning scheme (SCLS) for online missing PMU data correction (MPDC) in challenging measurement contexts, showing high efficacy in refining correction results and filtering out potential noises.
IEEE INTERNET OF THINGS JOURNAL
(2021)
Article
Environmental Sciences
Nan Wang, Xiaoling Zhang, Tianwen Zhang, Liming Pu, Xu Zhan, Xiaowo Xu, Yunqiao Hu, Jun Shi, Shunjun Wei
Summary: Phase filtering is a crucial step in InSAR terrain elevation measurements. Existing methods can be divided into traditional model-based and deep learning-based approaches, with deep learning methods often outperforming traditional ones. However, most existing deep learning methods rely heavily on complex architectures and large-scale training sets. In this study, we propose a sparse-model-driven network (SMD-Net) that models the physical filtering process in the network with fewer layers and parameters. The SMD-Net incorporates a convolutional neural network module to adaptively learn the sparse transform, significantly improving filtering performance. Experimental results on simulated and measured data demonstrate that the proposed method surpasses advanced InSAR phase filtering methods in terms of accuracy and speed. Furthermore, even with a small number of training samples, the proposed method still performs comparably on simulated data and outperforms another deep learning-based method on real data, indicating that it is not constrained by the requirement for a large number of training samples.
Article
Chemistry, Multidisciplinary
Shima Tavakoli, Nithiyanandan Krishnan, Hamidreza Mokhtari, Oommen P. Oommen, Oommen P. Varghese
Summary: A novel extracellular matrix-based bioink for stem cell 3D bioprinting is developed, which exhibits fast gelation kinetics, shear-thinning and shape-maintaining properties. The bioink significantly improves cell survival, stemness marker expression, cell proliferation, and migration. The dual cross-linking of the bioink contributes to self-healing, long-term stability, and enhanced cell proliferation.
ADVANCED FUNCTIONAL MATERIALS
(2023)
Article
Engineering, Electrical & Electronic
Zuhao Liu, Huan Wang, Yibo Gao, Shunchen Shi
Summary: This article introduces a new attention learning method based on NAS for detecting cardiovascular diseases, which outperforms existing methods in experiments.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Engineering, Aerospace
Ahmad Mahphood, Hossein Arefi
ADVANCES IN SPACE RESEARCH
(2020)
Article
Engineering, Aerospace
Anahita Asadi, Hossein Arefi, Hafez Fathipoor
ADVANCES IN SPACE RESEARCH
(2020)
Article
Environmental Sciences
Mehdi Khoshboresh-Masouleh, Fatemeh Alidoost, Hossein Arefi
JOURNAL OF APPLIED REMOTE SENSING
(2020)
Article
Chemistry, Analytical
Shima Sahebdivani, Hossein Arefi, Mehdi Maboudi
Review
Forestry
Vahid Nasiri, Ali A. Darvishsefat, Hossein Arefi, Marc Pierrot-Deseilligny, Manochehr Namiranian, Arnaud Le Bris
Summary: This research successfully demonstrated the potential of using low-cost UAV aerial images to estimate tree height and crown diameter, with high agreement between estimates and field measurements. The results confirmed the accuracy and feasibility of this approach for estimating tree heights and crown diameter.
CANADIAN JOURNAL OF FOREST RESEARCH
(2021)
Article
Environmental Sciences
Reza Akbari Dotappeh Sofla, Tayeb Alipour-Fard, Hossein Arefi
Summary: Road extraction, an important research topic in the fields of traffic management, road monitoring, and autonomous driving cars, has been addressed using a deep learning method based on U-net and SE. The method effectively recognizes roads and outperforms other networks in testing on two road datasets.
JOURNAL OF APPLIED REMOTE SENSING
(2021)
Article
Remote Sensing
Hamed Amini Amirkolaee, Hossein Arefi
Summary: This paper proposes a method to generate DSM using CNN, which successfully extracts features and creates a digital surface model through deep CNN structure and filters. The final integrated DSM shows high accuracy.
REMOTE SENSING LETTERS
(2021)
Article
Geochemistry & Geophysics
T. Alipour-Fard, M. E. Paoletti, Juan M. Haut, H. Arefi, J. Plaza, A. Plaza
Summary: The traditional design of CNNs overlooks the challenge of different scale features in HSI classification problems, while the newly developed MSKNet achieves modeling of different scales through multiple branches and attention mechanism, outperforming state-of-the-art CNNs in the context of HSI classification problems.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2021)
Article
Geochemistry & Geophysics
Nurollah Tatar, Hossein Arefi, Michael Hahn
Summary: The proposed method utilizes superpixels for object-based stereo matching, considering homogeneity weight for cost filtering and weighted cost aggregation by image objects. The iterative guided edge-preserving filter refines the disparity map significantly, improving the stereo matching result.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2021)
Article
Environmental Sciences
Mansour Mehranfar, Hossein Arefi, Fatemeh Alidoost
Summary: This study proposes a knowledge-based approach for automatic 3D reconstruction of main bridge elements such as railing, body, piers, and abutment using dense point clouds of unmanned aerial vehicle images. The method relies on geometric relations between bridge elements and has been shown to be effective in generating 3D models of different bridges.
JOURNAL OF APPLIED REMOTE SENSING
(2021)
Article
Environmental Sciences
Hamed Amini Amirkolaee, Hossein Arefi, Mohammad Ahmadlou, Vinay Raikwar
Summary: In this paper, a deep learning-based approach is proposed for directly generating DTM from DSM in complex scenes. The data is preprocessed and a hybrid deep convolutional neural network (HDCNN) is used for DTM extraction. A multi-scale fusion strategy is applied to generate the final DTM. The experiment results demonstrate significant performance and high generalizability of the proposed approach.
REMOTE SENSING OF ENVIRONMENT
(2022)
Article
Computer Science, Interdisciplinary Applications
Hamed Afsharnia, Hossein Arefi, Madjid Abbasi
Summary: This paper introduces a bias correction method for RPCs of satellite stereo images without GCPs, using global DEMs as control information and developing new formulae given directly in geodetic longitude and latitude format instead of Cartesian map projection coordinates. Experiment results show significant improvement in geopositioning accuracy and RMS improvement in longitude, latitude, and height.
EARTH SCIENCE INFORMATICS
(2022)
Article
Environmental Sciences
Vahid Nasiri, Ali Asghar Darvishsefat, Hossein Arefi, Verena C. Griess, Seyed Mohammad Moein Sadeghi, Stelian Alexandru Borz
Summary: This study successfully modeled forest canopy cover in the Hyrcanian mixed temperate forest in Northern Iran using a combination of Sentinel-2 data, high-resolution aerial images, and machine learning algorithms. The results showed that vegetation indices were the most important predictors in the models, and the random forest algorithm performed the best while the elastic net algorithm performed the worst in terms of model performance.
Article
Construction & Building Technology
Ahmad Mahphood, Hossein Arefi
Summary: In this paper, a novel method is proposed for extracting building boundaries from gridded building points. The method involves the generation of a grid, extraction of boundary points, and the use of a propagation algorithm for tracing and enlarging the outline. The results show a significant improvement in accuracy, with up to 70% improvement achieved using the proposed method.
AUTOMATION IN CONSTRUCTION
(2022)
Article
Chemistry, Analytical
Ali Abdollahi, Hossein Arefi, Shirin Malihi, Mehdi Maboudi
Summary: This paper focuses on 3D modeling of interior spaces in buildings using three-dimensional point clouds obtained from laser scanners. The walls, ceiling, and floor are extracted as the main structural fabric and reconstructed. The paper presents a method to address data-related issues such as obstruction, clutter, and noise. By employing a model-driven approach using watertight predefined models, the algorithm is able to effectively reconstruct non-rectangular spaces with an even number of sides. The proposed method is evaluated using real-world datasets, demonstrating its effectiveness in terms of completeness, correctness, and geometric accuracy with values ranging between [77%, 95%], [85%, 97%], and [1.7 cm, 2.4 cm], respectively.