Article
Computer Science, Software Engineering
Cheng-ming Liu, Wan-na Luan, Rong-hua Fu, Hai-bo Pang, Ying-hao Li
Summary: In this paper, a novel attention-embedding strategy for 3D saliency estimation is proposed by directly applying the attention embedding scheme to 3D mesh. The network can be trained in a weakly supervised manner and generalize well on different categories of objects. Experimental results demonstrate the effectiveness of the proposed method and its applicability to mesh simplification.
Article
Computer Science, Software Engineering
Rafael Kuffner dos Anjos, Richard Andrew Roberts, Benjamin Allen, Joaquim Jorge, Ken Anjyo
Summary: Highly complex and dense models of 3D objects are crucial in digital industries. Mesh decimation is important for efficiently simplifying complex meshes, and a preferred approach is to detect saliency and allow artists to iterate before simplification. We propose an efficient multi-scale method to compute mesh saliency, ensuring robust calculation even for densely tessellated models. Our implementation achieves significant speedups and is applicable to real use-case scenarios.
COMPUTERS & GRAPHICS-UK
(2023)
Article
Mathematical & Computational Biology
Stano Pekar, Jonas O. Wolff, Ludmila Cernecka, Klaus Birkhofer, Stefano Mammola, Elizabeth C. Lowe, Caroline S. Fukushima, Marie E. Herberstein, Adam Kucera, Bruno A. Buzatto, El Aziz Djoudi, Marc Domenech, Alison Vanesa Enciso, Yolanda M. G. Pinanez Espejo, Sara Febles, Luis F. Garcia, Thiago Goncalves-Souza, Marco Isaia, Denis Lafage, Eva Liznarova, Nuria Macias-Hernandez, Ivan Magalhaes, Jagoba Malumbres-Olarte, Ondrej Michalek, Peter Michalik, Radek Michalko, Filippo Milano, Ana Munevar, Wolfgang Nentwig, Giuseppe Nicolosi, Christina J. Painting, Julien Petillon, Elena Piano, Kaina Privet, Martin J. Ramirez, Candida Ramos, Milan Rezac, Aurelien Ridel, Vlastimil Ruzicka, Irene Santos, Lenka Sentenska, Leilani Walker, Kaja Wierucka, Gustavo Andres Zurita, Pedro Cardoso
Summary: Spiders, as ubiquitous predators in terrestrial ecosystems, play an important role and are suitable for testing eco-evolutionary hypotheses. An online database has been developed to archive and access diverse spider traits data globally, curated by an expert team and aimed at facilitating cross-taxon assays in functional ecology and comparative biology.
DATABASE-THE JOURNAL OF BIOLOGICAL DATABASES AND CURATION
(2021)
Article
Computer Science, Information Systems
Haoyang Xie, Yueqi Zhong
Summary: This paper aims to achieve high-performance, consistent 3D human mesh segmentation without requiring massive labelled data and tedious training. It proposes a Laplacian operator incorporating mesh saliency and a combinatorial descriptor derived from the spectrum of the saliency Laplacian operator to achieve consistent segmentation in the spectral domain. Experimental results demonstrate the effectiveness and efficiency of the proposed method for various 3D meshes, especially for human body shapes, without the need for time-consuming labelling and training on large-scale datasets.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Software Engineering
Baptiste Nicolet, Alec Jacobson, Wenzel Jakob
Summary: Inverse reconstruction from images is a challenging problem due to the non-convex nature of the objective function. Regularization techniques are commonly used to improve the robustness of optimization, but they come with their own set of issues. The proposed method introduces a preconditioned gradient descent approach to bias gradient steps towards smooth solutions, allowing for faster convergence without sacrificing the final solution's smoothness.
ACM TRANSACTIONS ON GRAPHICS
(2021)
Article
Computer Science, Software Engineering
Surya Kant Singh, Rajeev Srivastava
Summary: The method utilizes RGBD depth information for salient object detection, accurately and robustly detecting salient objects in complex and cluttered backgrounds, and producing a global concave reference surface.
Article
Chemistry, Multidisciplinary
Ning Xi, Yinjie Sun, Lei Xiao, Gang Mei
Summary: In this paper, a parallel adaptive Laplacian smoothing algorithm was proposed to improve the quality of large-scale tetrahedral meshes by utilizing the parallelism features of modern GPUs. The adaptive algorithm evaluates the mesh quality after each iteration using aspect ratio metrics to ensure improvement. Experimental tests showed that the proposed adaptive algorithm is significantly faster and improves accuracy, making it more applicable in dealing with tetrahedrons with poor quality.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Artificial Intelligence
Xiaoying Ding, Zhenzhong Chen
Summary: In this work, a novel 6DoF mesh saliency database is developed, providing both the subject's 6DoF data and eye-movement data. The inter-observer variation, influence of viewing direction, visual attention bias, and head movement during observation are analyzed based on the database. A 6DoF mesh saliency detection algorithm is proposed, taking into consideration uniqueness measure and bias preference. The experimental results demonstrate the superior performance of the approach, providing benchmarks for the presented 6DoF mesh saliency database.
Article
Mathematics, Applied
Xiu Ye, Shangyou Zhang
Summary: This article introduces a conforming discontinuous Galerkin (CDG) C-0-P-k finite element method for solving the biharmonic equation on triangular and tetrahedral meshes. The CDG method is based on taking weak divergence on the gradient of C-0-P-k finite elements, and optimal order error estimates in both a discrete H-2 norm and the L-2 norm are established.
APPLIED NUMERICAL MATHEMATICS
(2022)
Article
Computer Science, Artificial Intelligence
Ce Li, Shuxing Xuan, Fenghua Liu, Enbing Chang, Hailei Wu
Summary: This paper proposes a global attention network for cosaliency detection, which extracts and expresses collaborative saliency cues through the feature enhancement module (FEM), global information module (GIM), and collaboration correlation module (CCM). Experimental results demonstrate the effectiveness of the proposed method.
INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS
(2023)
Article
Computer Science, Theory & Methods
Xiaodong Wang, Fengju Kang, Hao Gu, Hongtao Liang
Summary: This paper proposes a mesh saliency detection algorithm considering both geometrical and colorimetric information. By computing local curvature entropy, multiple viewpoint projections, and other methods, salient regions on 3D meshes can be detected. Experiments show that the proposed method outperforms classic methods in terms of performance.
INTERNATIONAL JOURNAL OF MODELING SIMULATION AND SCIENTIFIC COMPUTING
(2022)
Article
Computer Science, Interdisciplinary Applications
Shuo Li, Xinhua Lu
Summary: The article proposes a novel second-order monotone upstream scheme for conservation law (MUSCL) scheme on unstructured quadrilateral meshes to solve the depth-averaged 2D nonlinear shallow-water equations involving wet-dry transitions on uneven bed terrains. The proposed scheme utilises the geometry of quadrilateral cells and is similar to the MUSCL scheme on a Cartesian mesh. The scheme includes a physically-based criterion to reduce the scheme to first-order in problematic cells near wet-dry interfaces and does not require variable gradients at cell-centers. Numerical tests demonstrate the robustness and accuracy of the developed numerical model.
INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN FLUIDS
(2023)
Article
Computer Science, Artificial Intelligence
Ya Zhang, Chunyi Chen, Xiaojuan Hu, Ling Li, Hailan Li
Summary: This paper introduces a novel saliency detection method that captures salient regions in textured models by fusing saliency maps from multi-view projections. A texel descriptor is introduced and a dataset reflecting human eye fixation patterns is created as an evaluation metric. Experimental results demonstrate that the proposed method outperforms existing approaches on multiple evaluation metrics.
PEERJ COMPUTER SCIENCE
(2023)
Article
Mathematics, Interdisciplinary Applications
Eylem Ozturk, Joseph L. Shomberg
Summary: In this study, a viscous Cahn-Hilliard phase-separation model with memory and a nonlocal fractional Laplacian operator in the chemical potential is examined. The existence of global weak solutions is proven using a Galerkin approximation scheme. Continuous dependence estimate provides uniqueness of weak solutions and existence of a compact connected global attractor in the weak energy phase space.
FRACTAL AND FRACTIONAL
(2022)
Article
Computer Science, Information Systems
Hongwei Zhao, Jiaxin Wu, Danyang Zhang, Pingping Liu
Summary: The article introduces two modules to obtain better image descriptors for image retrieval, and combines them to obtain more representative image feature descriptors.
ISPRS INTERNATIONAL JOURNAL OF GEO-INFORMATION
(2021)
Article
Computer Science, Software Engineering
Ran Song, Yonghuai Liu, Paul L. Rosin
Summary: A novel network trained in a weakly supervised manner, called CfS-CNN, has been developed to solve the difficulty of collecting vertex-level annotation for mesh saliency detection. This network outperforms existing state-of-the-art methods in mesh saliency detection and can be directly applied to scene saliency. Experimental results demonstrate the significant improvement of this approach.
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
(2021)
Article
Computer Science, Software Engineering
Lingchen Dai, Kang Zhang, Xianjun Sam Zheng, Ralph R. Martin, Yina Li, Jinhui Yu
Summary: Understanding how people perceive the visual complexity of shapes has theoretical and practical implications, with one school focusing on local features and another emphasizing global characteristics. Inspired by neuroscience, a model incorporating both local and global features is 92% in agreement with human ratings. It is also consistent with the hierarchical perceptual learning theory on how different layers of neurons contribute to the perception of visual shape complexity.
Article
Computer Science, Information Systems
Zhongwei Zhao, Ran Song, Qian Zhang, Peng Duan, Youmei Zhang
Summary: In this paper, a joint training framework called JoT-GAN is proposed to simultaneously train GAN and person re-identification (re-id) model. The optimal solutions for both generator and re-id model are achieved through mutual guidance by a discriminator. Experimental results on benchmark datasets show that the proposed joint training framework outperforms existing methods with separate training and achieves state-of-the-art re-id performance.
ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Weidong Zhang, Youmei Zhang, Ran Song, Ying Liu, Wei Zhang
Summary: In this paper, a novel weakly supervised learning framework is proposed to effectively learn the 3D layout of an indoor scene using 2D layout segmentation mask as supervision. A deep neural network is employed to predict the plane parameters and camera intrinsic parameters, which are used to generate the 3D layout, depth map, and 2D segmentation. Label consistency and depth consistency are utilized as key objectives to supervise the learning process, while 3D geometric reasoning and prior knowledge are incorporated to ensure the realism and reasonableness of the learned 3D layout.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Robotics
Jiapeng Sheng, Yanyun Chen, Xing Fang, Wei Zhang, Ran Song, Yu Zheng, Yibin Li
Summary: This paper explores the locomotion mechanisms of mammals and applies them to the design of control architectures for robots. By incorporating a rhythm generator to activate periodic motor patterns and using joint position increments as the outputs of RL policy, the proposed method can achieve a full spectrum of quadruped locomotion in both simulated and real-world scenarios.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Likun Xia, Hao Zhang, Yufei Wu, Ran Song, Yuhui Ma, Lei Mou, Jiang Liu, Yixuan Xie, Ming Ma, Yitian Zhao
Summary: The vessel-like structure in biomedical images is essential for understanding diseases and diagnosing and treating them. However, existing segmentation methods often produce unsatisfactory results due to challenges in identifying crisp edges.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Automation & Control Systems
Shuo Yang, Shuai Song, Shilei Chu, Ran Song, Jiyu Cheng, Yibin Li, Wei Zhang
Summary: This study investigates how to effectively integrate heuristics into the online 3D bin-packing problem by combining heuristic methods with deep reinforcement learning (DRL) techniques. Three different heuristics are designed based on real-world physical rules and experiences of human packers. They are then modeled into the DRL framework, resulting in a novel heuristic DRL method. Experimental results demonstrate that this method achieves state-of-the-art bin packing performance and can reliably accomplish bin packing tasks in real logistics scenarios.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Zhiheng Li, Xuyuan Gong, Ran Song, Peng Duan, Jun Liu, Wei Zhang
Summary: This paper proposes a self and mutual adaptive matching (SMAM) module to enhance the discriminability of feature maps extracted from sparse skeleton representation for few-shot action recognition. Experimental results demonstrate that the proposed SMAM-Net outperforms other baselines on the large-scale NTU RGB + D 120 dataset and generalizes well on smaller datasets. Codes and pretrained models will be publicly available.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Interdisciplinary Applications
Hao Zhang, Ran Song, Liping Wang, Lin Zhang, Dawei Wang, Cong Wang, Wei Zhang
Summary: This paper proposes a local-to-global graph neural network (LG-GNN) to address the issue of accurate classification in brain disorders by considering non-imaging information and the relationship between subjects. The proposed LG-GNN achieves state-of-the-art performance in ASD and AD classification.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Automation & Control Systems
Shuo Yang, Wei Zhang, Ran Song, Weizhi Lu, Hesheng Wang, Yibin Li
Summary: Demonstration understanding is crucial for robot imitation learning. This work investigates the visual change-based representation and builds imitation learning pipelines in both explicit and implicit ways. Experimental results show that the visual change-based approaches achieve state-of-the-art performance and indicate the superiority of the implicit method for imitation learning.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2022)
Article
Engineering, Civil
Qian Zhang, Mingxin Zhang, Jinghe Liu, Xuanyu He, Ran Song, Wei Zhang
Summary: This paper proposes a simple but effective method to solve the hard-positive problem in vessel re-identification. It constructs a multi-level contrastive learning framework trained with a specifically designed intra-batch cluster-level contrastive loss and an instance-level contrastive loss. Experimental results on a newly proposed dataset show that this method achieves state-of-the-art performance compared to other unsupervised methods.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Automation & Control Systems
Shuo Yang, Wei Zhang, Ran Song, Jiyu Cheng, Hesheng Wang, Yibin Li
Summary: This article proposes a watch-and-act imitation learning pipeline that enables robots to learn diverse manipulations from visual demonstrations. The system consists of a captioning module for understanding the demonstration videos and a manipulation module for learning the demonstrated manipulations. Extensive experiments validate the superiority of the two modules and a real robotic arm is used for demonstrating the whole robotic imitation system.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2023)
Article
Computer Science, Information Systems
Lin Zhang, Mingxin Zhang, Ran Song, Ziying Zhao, Xiaolei Li
Summary: This paper proposes an unsupervised method based on graph convolutional network for embedding learning, which encodes the structural information between samples and leverages augmentation to learn discriminative embeddings.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Computer Science, Artificial Intelligence
Ran Song, Wei Zhang, Yitian Zhao, Yonghuai Liu, Paul L. Rosin
Summary: This paper proposes a framework that combines a Generative Adversarial Network and a Conditional Random Field to learn the visual saliency of 3D objects and scenes. The experimental results demonstrate that this method outperforms the existing approaches in predicting human fixations and addresses the research question about 3D visual saliency.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Automation & Control Systems
Wenhao Tan, Xing Fang, Wei Zhang, Ran Song, Teng Chen, Yu Zheng, Yibin Li
Summary: This paper proposes a learning system for quadruped robots that can adapt to various terrains without pre-training. By introducing a hierarchical framework and reinforcement learning, the robot achieves agile locomotion in the real world.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
