Article
Engineering, Aerospace
Yuanliang Xue, Guodong Jin, Tao Shen, Lining Tan, Lianfeng Wang
Summary: This paper addresses the problem of visual object tracking for Unmanned Aerial Vehicles (UAVs). A simple yet efficient tracker following the basic architecture of the Siamese neural network is proposed, which improves the classification ability from three stages. Experimental results demonstrate that the proposed tracker achieves favorable tracking performances in aerial benchmarks beyond 41 frames/s.
CHINESE JOURNAL OF AERONAUTICS
(2023)
Article
Automation & Control Systems
Pengwen Xiong, Junjie Liao, MengChu Zhou, Aiguo Song, Peter X. Liu
Summary: This article proposes a deeply supervised subspace learning method to help robots understand and perceive an object's properties during noncontact robot-object interaction. It extracts contactless feature information from noncontact sensors to retrieve cross-modal information, estimating and inferring material properties of known and unknown objects. Experimental results show the effectiveness of this approach compared to other advanced methods.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Computer Science, Information Systems
G. Madhu, A. Govardhan, Vinayakumar Ravi, Sandeep Kautish, B. Sunil Srinivas, Tanupriya Chaudhary, Manoj Kumar
Summary: An automated, fast, and accurate diagnostic model for one-shot detection and classification of malaria thin blood smears was developed using the Deep Siamese Capsule Network (D-SCN). Experimental results showed high detection accuracy and classification accuracy for the model.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Mingyong Li, Qiqi Li, Yan Ma, Degang Yang
Summary: With the development of mobile Internet technology and the popularization of smart devices, the amount of multimedia data has increased significantly. Cross-modal retrieval, which allows users to search for information across different modalities, has received extensive research attention. This paper proposes a novel method called Semantic-guided Autoencoder Adversarial Hashing (SAAH) for cross-modal retrieval. SAAH maximizes the semantic relevance of instances and maintains the immutability of cross-modal under the guidance of semantic multi-labels. Experimental results on widely used cross-modal datasets demonstrate the effectiveness of SAAH in achieving leading retrieval performance.
COMPLEX & INTELLIGENT SYSTEMS
(2022)
Article
Environmental Sciences
Zhiying Cao, Wenhui Diao, Xian Sun, Xiaode Lyu, Menglong Yan, Kun Fu
Summary: The study introduces an efficient C3Net model for semantic segmentation of multi-modal remote sensing images, striking a balance between speed and accuracy. By utilizing backbone networks and plug-and-play modules, it effectively extracts and recalibrates multi-modal features, while reducing the number of model parameters by redesigning the semantic contextual extraction module based on lightweight convolutional groups. Additionally, a multi-level knowledge distillation strategy is proposed to enhance the performance of the compact model.
Article
Computer Science, Information Systems
Lingyun Song, Xuequn Shang, Chen Yang, Mingxuan Sun
Summary: Cross-Modal Zero-Shot Hashing is an important image retrieval technique, and AG-MIH is a network designed for CMZSH. By learning instance-level hash codes and attributes, AG-MIH achieves state-of-the-art performance in both cross-modal and single-modal zero-shot image retrieval tasks.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Engineering, Electrical & Electronic
Zhonghua Xie, Lingjun Liu, Cheng Wang
Summary: This paper proposes a novel model-guided boosting framework for improving the restoration quality of image denoising methods. The framework can be flexibly extended to composite denoising and utilizes both external and internal image properties through the use of deep neural networks and low-rank regularization.
Article
Cell Biology
Winston Koh, Shawn Hoon
Summary: This study explores the use of deep metric learning methods to achieve transfer of cell type annotation across different scRNA-seq platforms, batches, and species, using only a small training set, achieving high accuracy.
FRONTIERS IN CELL AND DEVELOPMENTAL BIOLOGY
(2021)
Article
Chemistry, Analytical
Felix Marattukalam, Waleed Abdulla, David Cole, Pranav Gulati
Summary: The need for contactless vascular biometric systems has led to the development of a low-cost end-to-end contactless wrist vein biometric recognition system based on deep learning. A U-Net CNN structure was trained using the FYO wrist vein dataset to effectively extract and segment wrist vein patterns, achieving a Dice Coefficient of 0.723 for the extracted images. A CNN and Siamese Neural Network were utilized to match wrist vein images, resulting in the highest F1-score of 84.7%. The average matching time was less than 3 seconds on a Raspberry Pi. All subsystems were integrated with a designed GUI to form a fully functional deep learning-based wrist biometric recognition system.
Article
Computer Science, Artificial Intelligence
Yi Zhang, Guixi Liu, Hanlin Huang, Ruke Xiong
Summary: This paper proposes a unified tracking framework that combines lightweight Siamese network and template-guided learning to address the balance between high performance and real-time execution on resource-constrained devices. By using a two-step pruning method to compress the Siamese network and constructing a template-guided learning model, the proposed method achieves efficient tracking performance.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Oncology
Ilyass Moummad, Cyril Jaudet, Alexis Lechervy, Samuel Valable, Charlotte Raboutet, Zamila Soilihi, Juliette Thariat, Nadia Falzone, Joelle Lacroix, Alain Batalla, Aurelien Corroyer-Dulmont
Summary: Due to long waiting times for MRI, algorithms such as resampling and denoising models are developed to speed up image acquisition time. However, the impact of these algorithms on radiomics has been poorly studied. This study aimed to develop resampling and denoising DL models and evaluate their impact on radiomics. The findings showed that DL models were able to reconstruct low resolution and noised MRI images quickly into high quality images, restoring radiomic features.
Article
Computer Science, Artificial Intelligence
Shuang Xu, Jiangshe Zhang, Jialin Wang, Kai Sun, Chunxia Zhang, Junmin Liu, Junying Hu
Summary: This paper proposes a more interpretable network for guided image denoising, by building an observation model and using deep prior regularized optimization problem to design the network architecture. Experimental results show that the network outperforms several state-of-the-art methods in terms of performance.
INFORMATION FUSION
(2022)
Article
Engineering, Biomedical
Man-ni Chu, Ming -lei Yu, Jia-lien Hsu
Summary: This paper proposes a computational method for calculating the similarity between different languages or language varieties and validates its effectiveness through experiments. The results show that the proposed model outperforms comparative experiments in the identification task and can assist linguists in pre-classifying sound files.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2024)
Article
Automation & Control Systems
Jeonghoon Kwak, Kyon-Mo Yang, Ye Jun Lee, Min-Gyu Kim, Kap-Ho Seo
Summary: This paper proposes a method using few-shot learning with a Siamese network to track target-of-interest objects. By increasing the difference in features between the target object and non-target objects, the proposed method improves the accuracy of object recognition.
INTERNATIONAL JOURNAL OF CONTROL AUTOMATION AND SYSTEMS
(2023)
Article
Computer Science, Information Systems
Zhenqiu Shu, Yibing Bai, Donglin Zhang, Jun Yu, Zhengtao Yu, Xiao-Jun Wub
Summary: The proposed SCCGDH model aims to learn specific class centers and guide hashing learning, reducing intraclass variation of multimedia data. Experimental results demonstrate its superior performance on three cross-modal datasets compared to other state-of-the-art hashing approaches.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Information Systems
Shuai Liu, Shuai Wang, Xinyu Liu, Jianhua Dai, Khan Muhammad, Amir H. H. Gandomi, Weiping Ding, Mohammad Hijji, Victor Hugo C. de Albuquerque
Summary: Computer vision, particularly visual monitoring technology, has shown great potential in the complex monitoring environment. This article proposes a fuzzy inference-based monitoring method that utilizes human inertial thinking characteristics to infer the target's location and applies an alternative selection strategy based on thinking set. Experimental results on multiple datasets demonstrate the effectiveness and robustness of the proposed method in IoT-assisted monitoring.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Review
Computer Science, Theory & Methods
Fath U. Min Ullah, Mohammad S. Obaidat, Amin Ullah, Khan Muhammad, Mohammad Hijji, Sung Wook Baik
Summary: Recent advancements in intelligent surveillance systems for video analysis have attracted significant attention in the research community. Automatic violence detection systems using artificial neural networks and machine intelligence are in high demand in heavily crowded areas to ensure safety and security in smart cities. Extensive literature on violence detection has been published, but existing surveys are limited in scope. To address this, we conduct a comprehensive survey and analysis of the literature, examining machine learning strategies, neural network-based analysis, limitations, and datasets. We also discuss evaluation strategies, metrics, and provide recommendations for future research in violence detection.
ACM COMPUTING SURVEYS
(2023)
Editorial Material
Computer Science, Artificial Intelligence
Kit Yan Chan, Bilal Abu-Salih, Khan Muhammad, Vasile Palade, Rifai Chai
Article
Mathematics
Senthil Kumar Jagatheesaperumal, Khan Muhammad, Abdul Khader Jilani Saudagar, Joel J. P. C. Rodrigues
Summary: Fire accidents cause a high number of casualties and manually extinguishing the fire is risky. The development of fire-extinguishing robots with advanced functionalities is ongoing, however, early detection of fire is lacking in most systems. This study introduces a deep learning-based automatic fire extinguishing mechanism utilizing convolutional neural networks for fire detection and human presence in fire locations. Experimental results show that the best combination of neural network parameters is an Adam optimizer with softmax activation and a learning rate of 0.001. The proposed model was tested using a mobile robotic system in automatic and wireless control modes, successfully extinguishing fires.
Article
Engineering, Electrical & Electronic
Samrah Mehraj, Subreena Mushtaq, Shabir A. Parah, Kaiser J. Giri, Javaid A. Sheikh, Amir H. Gandomi, Mohammad Hijji, Brij B. Gupta, Khan Muhammad
Summary: Heritage multimedia is a valuable cultural asset that provides insights into earlier generations and their creative approach, lifestyle, and historical ideologies. It is also an important resource for boosting the local economy, sustainable communities, and tourism and business sectors. With the advancements in technology and 5G networks, protecting heritage media from unauthorized consumers is crucial. This study proposes a robust and blind watermarking-framework for cultural images (RBWCI) that uses the discrete cosine transform domain for ownership verification and copyright protection.
IEEE TRANSACTIONS ON CONSUMER ELECTRONICS
(2023)
Article
Computer Science, Hardware & Architecture
Shuai Liu, Xiyu Xu, Yang Zhang, Khan Muhammad, Weina Fu
Summary: This article introduces a reliable sample selection strategy for weakly supervised visual tracking and verifies its importance in improving model performance. Experiments demonstrate that a scientific sample quality assessment method is of great help to data-based weakly supervised learning systems.
IEEE TRANSACTIONS ON RELIABILITY
(2023)
Article
Computer Science, Information Systems
Dan Wang, Bo Li, Bin Song, Yingjie Liu, Khan Muhammad, Xiaokang Zhou
Summary: In this article, a novel blockchain-supported hierarchical digital twin IoT (HDTIoT) framework is proposed to achieve secure and reliable real-time computation. The framework combines digital twin with edge network and adopts blockchain technology. By utilizing a data and knowledge dual-driven learning solution, the communication and computation efficiency is improved. Experimental results demonstrate the efficiency and reliability of the proposed resource allocation scheme in the HDTIoT system.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Review
Engineering, Multidisciplinary
Muhammad Sajjad, Fath U. Min Ullah, Mohib Ullah, Georgia Christodoulou, Faouzi Alaya Cheikh, Mohammad Hijji, Khan Muhammad, Joel J. P. C. Rodrigues
Summary: Facial expression recognition (FER) is a complex research topic with applications in various fields, such as healthcare and security. Computational FER mimics human facial expression coding skills to assist human-computer interaction. This study thoroughly analyzes and surveys the existing literature on FER, highlights the working flow of FER methods, discusses limitations in existing surveys, investigates FER datasets, and comprehensively discusses measures to evaluate FER performance.
ALEXANDRIA ENGINEERING JOURNAL
(2023)
Article
Engineering, Electrical & Electronic
Samee Ullah Khan, Noman Khan, Tanveer Hussain, Khan Muhammad, Mohammad Hijji, Javier Del Ser, Sung Wook Baik
Summary: This article proposes a multi-scale pyramid attention model for person re-identification (P-ReID) that leverages the complementarity between semantic attributes and visual appearance. The proposed model consists of three steps, including individual training of backbone model and appearance/attribute networks, fusion of dual network features, and re-training for P-ReID.
IEEE JOURNAL OF SELECTED TOPICS IN SIGNAL PROCESSING
(2023)
Article
Engineering, Civil
Khan Muhammad, Hayat Ullah, Salman Khan, Mohammad Hijji, Jaime Lloret
Summary: This paper proposes an efficient and lightweight CNN architecture for early fire detection and segmentation. By utilizing depth-wise separable convolution, point-wise group convolution, and a channel shuffling strategy, the model size and computation costs are significantly reduced. Extensive experiments validate the effectiveness and robustness of the proposed method in fire segmentation.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Engineering, Civil
Zheng Wang, Yifeng Wang, Khan Muhammad
Summary: The purpose of this study is to investigate the pricing and deficiency of Network Car Hailing (NCH) Platform in Edge Computing (EC)-based Intelligent Transportation System. This study introduces EC to address the capacity and load balancing issues in the car-hailing platform, and constructs an EC-based online car-hailing resource allocation and pricing optimization model. Experimental results show that as the number of vehicles with computing tasks increases, the amount of resources purchased and the cost of paying also increase, while the utility function of NCH platforms and operators declines. The model constructed in this study can minimize average cost and energy consumption while maintaining low delay, providing reference for intelligent pricing and resource allocation in the later period of intelligent transportation.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Computer Science, Information Systems
Samee Ullah Khan, Ijaz Ul Haq, Noman Khan, Amin Ullah, Khan Muhammad, Huiling Chen, Sung Wook Baik, Victor Hugo C. de Albuquerque
Summary: The study proposes a cyber-physical system (CPS)-based person reidentification (P-ReID) framework for smart surveillance. The framework utilizes AI techniques and IoT environments to improve efficiency and overcome challenges in person reidentification. A dual attention dilated network (DADNet) and dual feature fusion method are introduced to enhance the person matching probability. Additionally, diversity orthogonality regularization is imposed on several CNN layers to boost the performance of DADNet. A comprehensive analysis and comparison demonstrate the strength of DADNet in AI-enabled IoT settings.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Article
Automation & Control Systems
Nayan Kumar Subhashis Behera, Pankaj Kumar Sa, Khan Muhammad, Sambit Bakshi
Summary: Person Re-identification (PRId) is essential for associating photographs/videos of individuals obtained from various occasions or across cameras, especially in emergencies. Part-level features play a crucial role in person retrieval, and using convolutional partition of body parts to learn discriminative features is highlighted in this research. The proposed method of Convolutional Part Refine (CPR) shows competitive performance and addresses the within-part inconsistency issue in partition strategies.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
Article
Computer Science, Hardware & Architecture
Khan Muhammad, Javier Del Ser, Naercio Magaia, Ramon Fonseca, Tanveer Hussain, Amir H. Gandomi, Mahmoud Daneshmand, Victor Hugo C. de Albuquerque
Summary: With the increasing popularity of smart devices and their need for data, edge computing and edge learning have become powerful tools. However, edge learning faces challenges such as latency sensitivity and resource consumption. This study proposes a prioritization framework for video data based on edge learning, which can reduce resource usage. Additionally, communication aspects related to edge learning are critically examined.