Article
Materials Science, Multidisciplinary
Claudio Zeni, Andrea Anelli, Aldo Glielmo, Kevin Rossi
Summary: This passage discusses the characteristics of machine learning potentials built upon high-dimensional atom-density representations, indicating that the probability density induced by training points in the representation space plays a significant role in robust extrapolation and accurate prediction.
Article
Physics, Fluids & Plasmas
Satya N. Majumdar, Francesco Mori, Hendrik Schawe, Gregory Schehr
Summary: This study computed the mean perimeter and mean area of the convex hull of Brownian motion with resetting, showing that the convex hull approaches a circular shape at late times. The analytical predictions were confirmed through numerical simulations.
Article
Operations Research & Management Science
Carlos Alegria, David Orden, Carlos Seara, Jorge Urrutia
Summary: This study focuses on the rectilinear convex hull of a set of points in the plane, presenting a method to compute the minimum area under different angles and their optimal time complexity. The sorting and analysis of lines are carried out based on the angle Theta.
JOURNAL OF GLOBAL OPTIMIZATION
(2021)
Article
Plant Sciences
Xiaobao Liu, Biao Xu, Wenjuan Gu, Yanchao Yin, Hongcheng Wang
Summary: This study proposes a plant leaf veins coupling feature representation and measurement method based on DeepLabV3+ to address the issues of slow segmentation, partial occlusion of leaf veins, and low measurement accuracy. By using a lightweight network and an improved algorithm, the proposed method achieves high segmentation accuracy and speed, and accurately measures the length and width of leaf veins.
FRONTIERS IN PLANT SCIENCE
(2022)
Article
Mathematics
Wenwen Liu, Shiu Yin Yuen, Kwok Wai Chung, Chi Wan Sung
Summary: In this paper, a new method for generating multidimensional convex landscapes is proposed using insights from computational geometry. The proposed generator is generic and does not prefer any specific analytical function.
Article
Mathematics
Senlin Wu, Yafang Lv, Keke Zhang, Chan He
Summary: Estimations of the covering functional of a convex body that can be expressed as the convex hull of a finite number of compact convex sets are presented, with a particular focus on providing an upper bound for the covering functional of zonotopes that can be written as the sum of n + k segments.
MATHEMATICAL INEQUALITIES & APPLICATIONS
(2022)
Article
Agriculture, Multidisciplinary
Kaixuan Zhao, Meng Zhang, Weizheng Shen, Xiaohang Liu, Jiangtao Ji, Baisheng Dai, Ruihong Zhang
Summary: A two-level model based on a 3D structure feature map, using convex hull distance, was constructed to improve the accuracy of automatic body condition scoring for dairy cows.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2023)
Article
Operations Research & Management Science
Alejandra Martinez-Moraian, David Orden, Leonidas Palios, Carlos Seara, Pawel Zylinski
Summary: This study focuses on the computation and maintenance of the O-Kernel within a polygon as the set of orientations O is rotated by an angle theta. Efficient algorithms are designed for cases of one or two orthogonal orientations and simple polygons, determining the intervals of theta where the O-Kernel is not empty and optimizing the area or perimeter at a specific theta value. The algorithms are further improved for simple orthogonal polygons, with results extended to cases where O is a set of multiple orientations.
JOURNAL OF GLOBAL OPTIMIZATION
(2021)
Article
Computer Science, Information Systems
Xiaohan Yuan, Shuyu Chen, Han Zhou, Chuan Sun, Lu Yuwen
Summary: In this paper, a novel and simple convex hull-based SMOTE (CHSMOTE) algorithm is proposed to overcome the weaknesses of SMOTE and alleviate class imbalance problem. CHSMOTE selects the border minority samples as initial samples, identifies the synthesis area based on convex hull, and generates more effective samples by enlarging the generation range. Extensive experiments demonstrate the effectiveness and superiority of the proposed algorithm.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Theory & Methods
Jakub Bielawski, Jacek Tabor
Summary: The paper introduces the notion of T-convex hull of fuzzy sets in t-norm fuzzy sets, proving the relationship between the T-convex hull and the convex hull as well as demonstrating two applications. It shows that the operation of forming T-convex hull behaves well with respect to algebraic operations and presents an analogue of the Shapley-Folkman theorem in the context of upper semicontinuous fuzzy subsets.
FUZZY SETS AND SYSTEMS
(2021)
Article
Automation & Control Systems
Zhuo Long, Mingjie Cai, Qingguo Li, Yizhu Li, Wanting Cai
Summary: Many extensions of rough sets have sought appropriate granular structures, but few have considered data-driven approaches to generating posets-structured coverings based on irregular granules. This study proposes a tree-structured model using norm granules obtained through an onion-peeling strategy. Comparative experiments show that CrossSift outperforms other methods in terms of dependency degree and classification accuracy, bridging the gap between rough sets and perceptrons, and contributing to dimensionality reduction, computer vision, and geometry.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Materials Science, Multidisciplinary
Anshuman Kumar, Zulfikhar A. Ali, Bryan M. Wong
Summary: Defects in materials significantly affect their performance, and calculating all possible defects in complex materials is computationally challenging. In this study, we successfully interfaced DFTB with CASM to enhance efficiency for calculating and pre-screening formation energies/convex hulls.
JOURNAL OF MATERIALS SCIENCE & TECHNOLOGY
(2023)
Article
Mathematics
Akos G. Horvath, Zsolt Langi
Summary: The aim of this study is to investigate the properties of the convex hull and the homothetic convex hull functions of a convex body in Euclidean n-space. It is found that the convex hull function does not uniquely determine the convex body. Additionally, the study proves the equivalence between the polar projection problem and the translative constant volume property of convex bodies, and provides a brief proof of some theorems regarding the homothetic convex hull function. Furthermore, the study applies the results to describe the properties of the illumination bodies of convex bodies.
GEOMETRIAE DEDICATA
(2022)
Article
Computer Science, Theory & Methods
Bin Pang
Summary: This paper introduces basic notions of (L, M)-fuzzy convex structures and explores their relationship with (L, M)-fuzzy convex structures, including equivalent forms, fuzzy hull operators, and fuzzy interval operators. It is shown that there is a Galois correspondence between (L, M)-fuzzy interval spaces and (L, M)-fuzzy convex spaces, with the category of arity 2 (L, M)-fuzzy convex spaces embedding as a fully reflective subcategory in the category of (L, M)-fuzzy interval spaces.
FUZZY SETS AND SYSTEMS
(2021)
Article
Biotechnology & Applied Microbiology
Zelong Liu, Alexander Zhou, Valentin Fauveau, Justine Lee, Philip Marcadis, Zahi A. Fayad, Jimmy J. Chan, James Gladstone, Xueyan Mei, Mingqian Huang
Summary: This study used deep learning to automatically measure knee anatomy and analyzed several key parameters. The results showed that the model accurately identified patellofemoral landmarks and produced human-comparable measurements, which can enhance our ability to analyze knee anatomy at scale and improve outcomes.
BIOENGINEERING-BASEL
(2023)
Article
Engineering, Electrical & Electronic
B. Joukovsky, P. Hu, A. Munteanu
ELECTRONICS LETTERS
(2020)
Article
Materials Science, Textiles
Pengpeng Hu, Nastaran Nourbakhsh, Jing Tian, Stephan Sturges, Vasile Dadarlat, Adrian Munteanu
TEXTILE RESEARCH JOURNAL
(2020)
Article
Computer Science, Artificial Intelligence
Remco Royen, Leon Denis, Quentin Bolsee, Pengpeng Hu, Adrian Munteanu
Summary: This study introduces a novel MaskLayer neural network layer and proposes a masked optimizer and balancing gradient rescaling approach to achieve quality scalability within the deep learning framework. Experimental results show that the cost of introducing scalability with MaskLayer remains limited.
Article
Engineering, Electrical & Electronic
P. Hu, A. Munteanu
Summary: The method presented is for registering 3D shapes without overlap, validated on the FAUST dataset for human body reconstruction, and it avoids iterative optimization compared to existing state-of-the-art methods.
ELECTRONICS LETTERS
(2021)
Article
Automation & Control Systems
Pengpeng Hu, Nastaran Nourbakhsh Kaashki, Vasile Dadarlat, Adrian Munteanu
Summary: Estimating 3D human body shape and pose under clothing is crucial for various applications, and existing methods face challenges due to expensive computation and lack of training data. The proposed Body PointNet is a learning-based approach that outperforms state-of-the-art methods in terms of accuracy and running time, by operating directly on raw point clouds and synthesizing dressed-human pseudoscans for training.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2021)
Article
Engineering, Biomedical
Kevin D. McCay, Pengpeng Hu, Hubert P. H. Shum, Wai Lok Woo, Claire Marcroft, Nicholas D. Embleton, Adrian Munteanu, Edmond S. L. Ho
Summary: The early diagnosis of cerebral palsy has been an important area of recent research. Automating diagnostic tools like General Movements Assessment (GMA) can improve accessibility and enhance understanding of infant movement development. This paper proposes new and improved features for classification of infant body movements using pose-based features extracted from RGB video sequences. The proposed framework shows good classification performance across multiple datasets.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2022)
Article
Engineering, Electrical & Electronic
Pengpeng Hu, Ran Zhao, Xinxin Dai, Adrian Munteanu
Summary: This paper introduces a deep learning approach for reconstructing a clean, watertight body mesh and normalizing the posture of the human body model from an input set of impaired body point clouds. The proposed method, named I2H, is the first deep learning approach to address these issues and achieves the best performance in 3D human mesh reconstruction.
Article
Engineering, Electrical & Electronic
Pengpeng Hu, Edmond S. L. Ho, Adrian Munteanu
Summary: This article proposes a novel deep learning framework for generating omnidirectional 3-D point clouds of human bodies by registering front- and back-facing partial scans. The method does not require calibration-assisting devices or assumptions on initial alignment or correspondences. The approach builds virtual correspondences for the partial scans and predicts the rigid transformation between them through deep neural networks. Experiments show that the proposed method achieves state-of-the-art performance in both objective and subjective terms.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Chemistry, Analytical
Xinxin Dai, Ran Zhao, Pengpeng Hu, Adrian Munteanu
Summary: Keystroke dynamics is a soft biometric based on unique typing patterns. The researchers explored a novel visual modality using a single RGB-D sensor and created a dataset called KD-MultiModal for human identification. The dataset contains RGB-D image sequences of individuals typing sentences and proved the effectiveness of the proposed method. The dataset is publicly accessible for further research.
Article
Engineering, Electrical & Electronic
Pengpeng Hu, Xinxin Dai, Ran Zhao, He Wang, Yingliang Ma, Adrian Munteanu
Summary: This study proposes a novel vision-based method called Point2PartVolume based on deep learning for rapidly and accurately predicting part-aware body volumes from a single-depth image of the dressed body. A multitask neural network is designed to complete partial body point clouds, predict body shape under clothing, and segment the reconstructed body into parts. Experimental results show that the proposed method outperforms the state-of-the-art with an average 90% volume prediction accuracy.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Computer Science, Information Systems
Nastaran Nourbakhsh Kaashki, Pengpeng Hu, Adrian Munteanu
Summary: 3D anthropometric measurement extraction is crucial for various applications. Existing methods often suffer from sensitivity to noise and missing data, as well as computational complexity. To address these limitations, we propose a deep neural network architecture that fits a template to the input scan and outputs reconstructed body and measurements, without the need for transferring and refining measurements. A novel loss function and two large datasets are introduced for training. Experimental results show that our approach outperforms state-of-the-art methods in terms of accuracy and robustness, using both synthesized and real 3D scans.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Nastaran Nourbakhsh, Xinxin Dai, Timea Gyarmathy, Pengpeng Hu, Bogdan Iancu, Adrian Munteanu
Summary: Accurate hand measurement data is crucial in medical science, fashion industry, and augmented/virtual reality applications. This paper introduces a deep-learning-based method to automatically measure the hand in a non-contact manner, outperforming existing methods in various hand measurement types.
PROCEEDINGS OF 2022 7TH INTERNATIONAL CONFERENCE ON MACHINE LEARNING TECHNOLOGIES, ICMLT 2022
(2022)
Proceedings Paper
Instruments & Instrumentation
Nastaran Nourbakhsh Kaashki, Xinxin Dai, Timea Gyarmathy, Pengpeng Hu, Bogdan Iancu, Adrian Munteanu
Summary: Recent advancements in 3D scanning technologies have enabled accurate extraction of hand geometry and measurements. In this paper, we propose the first deep neural network for automatic hand measurement extraction and train it with a novel synthetic dataset. Experimental results demonstrate the superiority of our method in terms of accuracy and speed compared to existing techniques.
2022 IEEE INTERNATIONAL INSTRUMENTATION AND MEASUREMENT TECHNOLOGY CONFERENCE (I2MTC 2022)
(2022)
Article
Computer Science, Information Systems
Pengpeng Hu, Edmond Shu-Lim Ho, Adrian Munteanu
Summary: This paper proposes a deep learning algorithm called 3DBodyNet for rapidly reconstructing the 3D shape of human bodies using a single commodity depth camera. The algorithm is easy to use and only requires two depth images, while being insensitive to pose variations.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Engineering, Electrical & Electronic
Nastaran Nourbakhsh Kaashki, Pengpeng Hu, Adrian Munteanu
Summary: The appearance of 3-D scanners has revolutionized anthropometric data collection systems, with 3-D-based methods becoming increasingly popular for anthropometric measurement extraction. The proposed method using deep-learning for automatic anthropometric measurements extraction outperforms state-of-the-art methods, showing promising results on synthetic and real data.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)