arXiv subjects

Hongyu Wu

Publications and source records attributed to Hongyu Wu.

At least 37 records · Page 2Linked to original sources

arXiv2025

Segment Any 3D-Part in a Scene from a Sentence

This paper aims to achieve the segmentation of any 3D part in a scene based on natural language descriptions, extending beyond traditional object-level 3D scene understanding and addressing both data and methodological challenges. Due to the expensive acquisition and annotation burden, existing datasets and methods are predominantly limited to object-level comprehension. To overcome the limitations of data and annotation availability, we introduce the 3D-PU dataset, the first large-scale 3D dataset with dense part annotations, created through an innovative and cost-effective method for constructing synthetic 3D scenes with fine-grained part-level annotations, paving the way for advanced 3D-part scene understanding. On the methodological side, we propose OpenPart3D, a 3D-input-only framework to effectively tackle the challenges of part-level segmentation. Extensive experiments demonstrate the superiority of our approach in open-vocabulary 3D scene understanding tasks at the part level, with strong generalization capabilities across various 3D scene datasets.

cs.CV↗

arXiv2025

3D Dental Model Segmentation with Geometrical Boundary Preserving

3D intraoral scan mesh is widely used in digital dentistry diagnosis, segmenting 3D intraoral scan mesh is a critical preliminary task. Numerous approaches have been devised for precise tooth segmentation. Currently, the deep learning-based methods are capable of the high accuracy segmentation of crown. However, the segmentation accuracy at the junction between the crown and the gum is still below average. Existing down-sampling methods are unable to effectively preserve the geometric details at the junction. To address these problems, we propose CrossTooth, a boundary-preserving segmentation method that combines 3D mesh selective downsampling to retain more vertices at the tooth-gingiva area, along with cross-modal discriminative boundary features extracted from multi-view rendered images, enhancing the geometric representation of the segmentation network. Using a point network as a backbone and incorporating image complementary features, CrossTooth significantly improves segmentation accuracy, as demonstrated by experiments on a public intraoral scan dataset.

cs.CV↗

arXiv2024

A compositional ordering-driven morphotropic phase boundary in ferroelectric solid solutions

Ferroelectric solid solutions usually exhibit giant dielectric response and high piezoelectricity in the vicinity of the morphotropic phase boundary (MPB), where the structural phase transitions between the rhombohedral and the tetragonal phases as a result of the composition or strain variation. Here, we propose a compositional ordering-driven MPB in the specified compositional solid solutions. By preforming machine-learning potential based molecular dynamics simulations on lead zirconate titanate, we find a phase transition from the rhombohedral to tetragonal phase with the decrease of compositional ordering, leading to the MPB on temperature-ordering phase diagram. The compositional ordering-driven MPB can enhances the piezoelectricity with a magnitude comparable to that at the composition-driven MPB. Finally, we demonstrate that the mechanism of high piezoelectricity is polarization rotation driven by external field. This work provides an additional degree of freedom, compositional ordering, to design the high-performance piezoelectric materials.

cond-mat.mtrl-sci↗

arXiv2024

Parametric Primitive Analysis of CAD Sketches with Vision Transformer

The design and analysis of Computer-Aided Design (CAD) sketches play a crucial role in industrial product design, primarily involving CAD primitives and their inter-primitive constraints. To address challenges related to error accumulation in autoregressive models and the complexities associated with self-supervised model design for this task, we propose a two-stage network framework. This framework consists of a primitive network and a constraint network, transforming the sketch analysis task into a set prediction problem to enhance the effective handling of primitives and constraints. By decoupling target types from parameters, the model gains increased flexibility and optimization while reducing complexity. Additionally, the constraint network incorporates a pointer module to explicitly indicate the relationship between constraint parameters and primitive indices, enhancing interpretability and performance. Qualitative and quantitative analyses on two publicly available datasets demonstrate the superiority of this method.

cs.CV↗

arXiv2024

FaceCom: Towards High-fidelity 3D Facial Shape Completion via Optimization and Inpainting Guidance

We propose FaceCom, a method for 3D facial shape completion, which delivers high-fidelity results for incomplete facial inputs of arbitrary forms. Unlike end-to-end shape completion methods based on point clouds or voxels, our approach relies on a mesh-based generative network that is easy to optimize, enabling it to handle shape completion for irregular facial scans. We first train a shape generator on a mixed 3D facial dataset containing 2405 identities. Based on the incomplete facial input, we fit complete faces using an optimization approach under image inpainting guidance. The completion results are refined through a post-processing step. FaceCom demonstrates the ability to effectively and naturally complete facial scan data with varying missing regions and degrees of missing areas. Our method can be used in medical prosthetic fabrication and the registration of deficient scanning data. Our experimental results demonstrate that FaceCom achieves exceptional performance in fitting and shape completion tasks. The code is available at https://github.com/dragonylee/FaceCom.git.

cs.CV↗

arXiv2023

Atomistic simulations of thermodynamic properties of liquid gallium from first principles

In the research of condensed matter, atomistic dynamic simulations play a crucial role, particularly in revealing dynamic processes, phase transitions and thermodynamic statistics macroscopic physical properties in systems such as solids and liquids. For a long time, simulating complex and disordered liquids has been a challenge compared to ordered crystalline structures. The primary reasons for this challenge are the lack of precise force field functions and the neglect of nuclear quantum effects. To overcome these two limits in simulation of liquids, we use a deep potential (DP) with quantum thermal bath (QTB) approach. DP is a machine learning model are sampled from density functional theory and able to do large-scale atomic simulations with its precision. QTB is a method which incorporates nuclear quantum effects by quantum fluctuation dissipation. The application of this first principles approach enable us to successfully describe the phase transition processes in solid and liquid Gallium (Ga) as well as the associated dynamic phenomena. More importantly, we obtain the thermodynamic properties of liquid Ga, such as internal energy, specific heat, enthalpy change, entropy and Gibbs free energy, and these results align remarkably well with experiments. Our research has opened up a new paradigm for the study of dynamics and thermodynamics in liquids, amorphous materials, and other disordered systems, providing valuable insights and references for future investigations.

cond-mat.mtrl-sci↗

arXiv2023

Pressure-induced one-dimensional oxygen ion diffusion channel in lead-apatite

Recently, Lee et al. claimed that the experimental observation of room-temperature ambient-pressure superconductivity in a Cu-doped lead-apatite (Pb10-xCux(PO4)6O). The study revealed the Cu doping induces a chemical pressure, resulting in a structural contraction of one-dimensional Cu-O-Cu atomic column. This unique structure promotes a one-dimensional electronic conduction channel along the c-axis mediated by the O atoms, which may be related to superconductivity. These O atoms occupy 1/4 of the equivalent positions along the c-axis and exhibit a low diffusion activation energy of 0.8 eV, indicating the possibility of diffusion between these equivalent positions. Here, using machine-learning based deep potential, we predict the pressure-induced fast diffusion of 1/4-occupied O atoms along the one-dimensional channel in Pb10(PO4)6O at 500 K, while the frameworks of Pb triangles and PO4 tetrahedrons remain stable. The calculation results also indicate Cu doping can provide appropriate effective chemical pressure, indicating the one-dimensional ion diffusion channel may appear in Pb9Cu(PO4)6O, even at ambient pressure. Our finding shows that the one-dimensional ions diffusion channel may be an important factor to affects the fabrication and electrical measurement of doped lead-apatite.

cond-mat.mtrl-sci↗

arXiv2023

Credit Information in Earnings Calls

We develop a novel technique to extract credit-relevant information from the text of quarterly earnings calls. This information is not spanned by fundamental or market variables and forecasts future credit spread changes. One reason for such forecastability is that our text-based measure predicts future credit spread risk and firm profitability. More firm- and call-level complexity increase the forecasting power of our measure for spread changes. Out-of-sample portfolio tests show the information in our measure is valuable for investors. Both results suggest that investors do not fully internalize the credit-relevant information contained in earnings calls.

q-fin.GN↗

arXiv2023

Zero-Point Quantum Diffusion of Proton in Hydrogen-rich Superconductor $LaH_{10}$

$LaH_{10}$, as a member of hydrogen-rich superconductors, has a superconducting critical temperature of 250 K at high pressures, which exhibits the possibility of solving the long-term goal of room temperature superconductivity. Considering the extreme pressure and low mass of hydrogen, the nuclear quantum effects in $LaH_{10}$ should be significant and have an impact on its various physical properties. Here, we adopt the method combines deep-potential (DP) and quantum thermal bath (QTB), which was verified to be able to account for quantum effects in high-accuracy large-scale molecular dynamics simulations. Our method can actually reproduce pressure-temperature phase diagrams of $LaH_{10}$ consistent with experimental and theoretical results. After incorporating quantum effects, the quantum fluctuation driven diffusion of proton is found even in the absence of thermal fluctuation near 0 K. The high mobility of proton is found to be compared to liquid, yet the structure of $LaH_{10}$ is still rigid. These results would greatly enrich our vision to study quantum behavior of hydrogen-rich superconductors.

cond-mat.mtrl-sci↗

arXiv2023

Links Assignment Scheme based on Potential Edges Importance in Dual-layer Wavelength Routing Optical Satellite Networks

With the development of the massive satellite constellation and the on-orbit laser-based communication equipment, the wavelength routing optical satellite network (WROSN) becomes a potential solution for on-orbit, high-capacity, and high-speed communication. Since the inter-satellite links (ISLs) are time-varying, one of the fundamental considerations in the construction of the WROSN is assigning limited laser communication terminals for each satellite to establish ISLs with the visible satellites. Therefore, we propose a links assignment scheme (LAS) based on the potential edges importance matrix (PEIM) algorithm to construct a temporarily stable topology of the ISLs for a dual-layer constellation. The simulation results showed that the LAS based on the PEIM algorithm is better than LAS based on the random or Greedy algorithm in terms of node-to-node distance, node pair connectivity, wavelength demand, and transmission delay. The node pair connectivity and wavelength demand in WROSN is a trade-off problem. The research in this paper also brings a novel method for reduction of the cost of the on-board resources, that is through designing topology of the ISLs with links assignment algorithm.

cs.NI↗

arXiv2022

Large-scale Atomistic Simulation of Quantum Effects in SrTiO$_3$ from First Principles

Quantum effects of lattice vibration play a major role in many physical properties of condensed matter systems, including thermal properties such as specific heat, structural phase transition, as well as phenomena such as quantum crystal and quantum paraelectricity that are closely related to zero-point fluctuations. However, realizing atomistic simulations for realistic materials with a fully quantum-mechanical description remains a great challenge. Here, we propose a first-principle strategy for large scale Molecular Dynamics simulation, where high accuracy force field obtained by Deep-Potential (DP) is combined with Quantum Thermal Bath (QTB) method to account for quantum effects. We demonstrate the power of this DP+QTB method using the archetypal example SrTiO$_3$, which exhibits several phenomena induced by quantum fluctuations, such as the suppressed structure phase transition temperature, the quantum paraelectric ground state at low temperature and the quantum critical behavior $1/T^2$ law of dielectric constant. Our DP+QTB strategy is efficient in simulating large scale system, and is first principle. More importantly, quantum effects of other systems could also be investigated as long as corresponding DP model is trained. This strategy would greatly enrich our vision and means to study quantum behavior of condensed matter physics.

cond-mat.mtrl-sci↗

arXiv2022

Large-scale atomistic simulation of dislocation core structure in face-centered cubic metal with Deep Potential method

The core structure of dislocations is critical to their mobility, cross slip, and other plastic behaviors. Atomistic simulation of the core structure is limited by the size of first-principles density functional theory (DFT) calculation and the accuracy of classical molecular dynamics with empirical interatomic potentials. Here, we utilize a Deep Potential (DP) method learned from DFT calculations to investigate the dislocations of face-centered cubic copper on a large scale and obtain their core structures and energies. The validity of the DP description of the core structure and elastic strain from dislocation is confirmed by a fully discrete Peierls model. Moreover, the DP method can be further extended easily to dislocations with defects such as surface or vacancy, and our study will pave a way in the large-scale atomistic simulation of dislocation on the DFT level.

cond-mat.mtrl-sci↗

arXiv2022

Origin of negative thermal expansion and pressure induced amorphization in zirconium tungstate from machine-learning potential

Understanding various macroscopic pressure-volume-temperature properties of materials on the atomistic level has always been an ambition for physicists and material scientists. Particularly, some materials such as zirconium tungstate (ZrW2O8), exhibit multiple exotic properties including negative thermal expansion (NTE) and pressure-induced amorphization (PIA). Here, using machine-learning based deep potential, we trace both of the phenomena in ZrW2O8 back to a common atomistic origin, where the nonbridging O atoms play a critical role. We demonstrate that the nonbridging O atoms confer great flexibility to vibration of polyhedrons, and kinetically drive volume shrinking on heating, or NTE. In addition, beyond a certain critical pressure, we find that the migration of nonbridging O atoms leads to additional bond formation that lowers the potential energy, suggesting that the PIA is a potential-driven first-order phase transition. Most importantly, we identify a second critical pressure beyond which the amorphous phase of ZrW2O8 undergoes a hidden phase transition from a reversible phase to an irreversible one.

cond-mat.mtrl-sci↗

arXiv2022

Structural Phase Transitions in SrTiO3 from Deep Potential Molecular Dynamics

Strontium titanate (SrTiO3) is regarded as an essential material for oxide electronics. One of its many remarkable features is subtle structural phase transition, driven by antiferrodistortive lattice mode, from a high-temperature cubic phase to a low-temperature tetragonal phase. Classical molecular dynamics (MD) simulation is an efficient technique to reveal atomistic features of phase transition, but its application is often limited by the accuracy of empirical interatomic potentials. Here, we develop an accurate deep potential (DP) model of SrTiO3 based on a machine learning method using data from first-principles density functional theory (DFT) calculations. The DP model has DFT-level accuracy, capable of performing efficient MD simulations and accurate property predictions. Using the DP model, we investigate the temperature-driven cubic-to-tetragonal phase transition and construct the in-plane biaxial strain-temperature phase diagram of SrTiO3. The simulations demonstrate that strain-induced ferroelectric phase is characterized by two order parameters, ferroelectric distortion and antiferrodistortion, and the ferroelectric phase transition has both displacive and order-disorder characters. This works lays the foundation for the development of accurate DP models of other complex perovskite materials.

cond-mat.mtrl-sci↗

arXiv2021

Charged Oxygen Vacancy Induced Ferroelectric Structure Transition in Hafnium Oxide

The discovery of ferroelectric HfO2 in thin films and more recently in bulk is an important breakthrough because of its silicon-compatibility and unexpectedly persistent polarization at low dimensions, but the origin of its ferroelectricity is still under debate. The stabilization of the metastable polar orthorhombic phase was often considered as the cumulative result of various extrinsic factors such as stress, grain boundary, and oxygen vacancies as well as phase transition kinetics during the annealing process. We propose a novel mechanism to stabilize the polar orthorhombic phase over the nonpolar monoclinic phase that is the bulk ground state. Our first-principles calculations demonstrate that the doubly positively charged oxygen vacancy, an overlooked defect but commonly presented in binary oxides, is critical for the stabilization of ferroelectric phase. The charge state of oxygen vacancy serves as a new degree of freedom to control the thermodynamic stability of competing phases of wide-band-gap oxides.

physics.comp-ph↗

arXiv2020

A Data-Driven Machine Learning Approach for Consumer Modeling with Load Disaggregation

While non-parametric models, such as neural networks, are sufficient in the load forecasting, separate estimates of fixed and shiftable loads are beneficial to a wide range of applications such as distribution system operational planning, load scheduling, energy trading, and utility demand response programs. A semi-parametric estimation model is usually required, where cost sensitivities of demands must be known. Existing research work consistently uses somewhat arbitrary parameters that seem to work best. In this paper, we propose a generic class of data-driven semiparametric models derived from consumption data of residential consumers. A two-stage machine learning approach is developed. In the first stage, disaggregation of the load into fixed and shiftable components is accomplished by means of a hybrid algorithm consisting of non-negative matrix factorization (NMF) and Gaussian mixture models (GMM), with the latter trained by an expectation-maximization (EM) algorithm. The fixed and shiftable loads are subject to analytic treatment with economic considerations. In the second stage, the model parameters are estimated using an L2-norm, epsilon-insensitive regression approach. Actual energy usage data of two residential customers show the validity of the proposed method.

eess.SP↗

arXiv2020

Autonomous Unknown-Application Filtering and Labeling for DL-based Traffic Classifier Update

Network traffic classification has been widely studied to fundamentally advance network measurement and management. Machine Learning is one of the effective approaches for network traffic classification. Specifically, Deep Learning (DL) has attracted much attention from the researchers due to its effectiveness even in encrypted network traffic without compromising neither user privacy nor network security. However, most of the existing models are created from closed-world datasets, thus they can only classify those existing classes previously sampled and labeled. In this case, unknown classes cannot be correctly classified. To tackle this issue, an autonomous learning framework is proposed to effectively update DL-based traffic classification models during active operations. The core of the proposed framework consists of a DL-based classifier, a self-learned discriminator, and an autonomous self-labeling model. The discriminator and self-labeling process can generate new dataset during active operations to support classifier update. Evaluation of the proposed framework is performed on an open dataset, i.e., ISCX VPN-nonVPN, and independently collected data packets. The results demonstrate that the proposed autonomous learning framework can filter packets from unknown classes and provide accurate labels. Thus, corresponding DL-based classification models can be updated successfully with the autonomously generated dataset.

cs.NI↗

arXiv2019

Robust Matrix Completion State Estimation in Distribution Systems

Due to the insufficient measurements in the distribution system state estimation (DSSE), full observability and redundant measurements are difficult to achieve without using the pseudo measurements. The matrix completion state estimation (MCSE) combines the matrix completion and power system model to estimate voltage by exploring the low-rank characteristics of the matrix. This paper proposes a robust matrix completion state estimation (RMCSE) to estimate the voltage in a distribution system under a low-observability condition. Tradition state estimation weighted least squares (WLS) method requires full observability to calculate the states and needs redundant measurements to proceed a bad data detection. The proposed method improves the robustness of the MCSE to bad data by minimizing the rank of the matrix and measurements residual with different weights. It can estimate the system state in a low-observability system and has robust estimates without the bad data detection process in the face of multiple bad data. The method is numerically evaluated on the IEEE 33-node radial distribution system. The estimation performance and robustness of RMCSE are compared with the WLS with the largest normalized residual bad data identification (WLS-LNR), and the MCSE.

cs.LG↗