arXiv subjects

Bowen Wang

Publications and source records attributed to Bowen Wang.

At least 109 records · Page 6Linked to original sources

arXiv2022

Panoptic-aware Image-to-Image Translation

Despite remarkable progress in image translation, the complex scene with multiple discrepant objects remains a challenging problem. The translated images have low fidelity and tiny objects in fewer details causing unsatisfactory performance in object recognition. Without thorough object perception (i.e., bounding boxes, categories, and masks) of images as prior knowledge, the style transformation of each object will be difficult to track in translation. We propose panoptic-aware generative adversarial networks (PanopticGAN) for image-to-image translation together with a compact panoptic segmentation dataset. The panoptic perception (i.e., foreground instances and background semantics of the image scene) is extracted to achieve alignment between object content codes of the input domain and panoptic-level style codes sampled from the target style space, then refined by a proposed feature masking module for sharping object boundaries. The image-level combination between content and sampled style codes is also merged for higher fidelity image generation. Our proposed method was systematically compared with different competing methods and obtained significant improvement in both image quality and object recognition performance.

cs.CV↗

arXiv2022

Automated Grading System of Retinal Arterio-venous Crossing Patterns: A Deep Learning Approach Replicating Ophthalmologist's Diagnostic Process of Arteriolosclerosis

The status of retinal arteriovenous crossing is of great significance for clinical evaluation of arteriolosclerosis and systemic hypertension. As an ophthalmology diagnostic criteria, Scheie's classification has been used to grade the severity of arteriolosclerosis. In this paper, we propose a deep learning approach to support the diagnosis process, which, to the best of our knowledge, is one of the earliest attempts in medical imaging. The proposed pipeline is three-fold. First, we adopt segmentation and classification models to automatically obtain vessels in a retinal image with the corresponding artery/vein labels and find candidate arteriovenous crossing points. Second, we use a classification model to validate the true crossing point. At last, the grade of severity for the vessel crossings is classified. To better address the problem of label ambiguity and imbalanced label distribution, we propose a new model, named multi-diagnosis team network (MDTNet), in which the sub-models with different structures or different loss functions provide different decisions. MDTNet unifies these diverse theories to give the final decision with high accuracy. Our severity grading method was able to validate crossing points with precision and recall of 96.3% and 96.3%, respectively. Among correctly detected crossing points, the kappa value for the agreement between the grading by a retina specialist and the estimated score was 0.85, with an accuracy of 0.92. The numerical results demonstrate that our method can achieve a good performance in both arteriovenous crossing validation and severity grading tasks. By the proposed models, we could build a pipeline reproducing retina specialist's subjective grading without feature extractions. The code is available for reproducibility.

eess.IV↗

arXiv2022

Double-Phase-Shifter based Hybrid Beamforming for mmWave DFRC in the Presence of Extended Target and Clutters

In millimeter-wave (mmWave) dual-function radar-communication (DFRC) systems, hybrid beamforming (HBF) is recognized as a promising technique utilizing a limited number of radio frequency chains. In this work, in the presence of extended target and clutters, a HBF design based on the subarray connection architecture is proposed for a multiple-input multiple-output (MIMO) DFRC system. In this HBF, the double-phase-shifter (DPS) structure is embedded to further increase the design flexibility. We derive the communication spectral efficiency (SE) and radar signal-to-interference-plus-noise-ratio (SINR) with respect to the transmit HBF and radar receiver, and formulate the HBF design problem as the SE maximization subjecting to the radar SINR and power constraints. To solve the formulated nonconvex problem, the joinT Hybrid bRamforming and Radar rEceiver OptimizatioN (THEREON) is proposed, in which the radar receiver is optimized via the generalized eigenvalue decomposition, and the transmit HBF is updated with low complexity in a parallel manner using the consensus alternating direction method of multipliers (consensus-ADMM). Furthermore, we extend the proposed method to the multi-user multiple-input single-output (MU-MISO) scenario. Numerical simulations demonstrate the efficacy of the proposed algorithm and show that the solution provides a good trade-off between number of phase shifters and performance gain of the DPS HBF.

eess.SP↗

arXiv2022

Phonon anharmonicity and thermal conductivity of two-dimensional van der Waals materials: A review

Two-dimensional (2D) van der Waals (vdW) materials have extraordinary thermal properties due to the effect of quantum confinement, making them promising for thermoelectric energy conversion and thermal management in microelectronic devices. In this review, the mechanism of phonon anharmonicity originating from three- and four-phonon interactions is derived. The phonon anharmonicity of 2D vdW materials, involving the Grüneisen parameter, phonon lifetime, and thermal conductivity, is summarized and derived in detail. The size-dependent thermal conductivity of representative 2D vdW materials is discussed experimentally and theoretically. This review will present fundamental and advanced knowledge on how to evaluate the phonon anharmonicity in 2D vdW materials, which will aid the design of new structures and materials for applications related to energy transfer and conversion.

cond-mat.mtrl-sci↗

arXiv2022

First-principles Study of Phonon Lifetime and Low Lattice Thermal Conductivity of Monolayer γ-GeSe: A Comparative Study

Germanium selenide (GeSe) is a unique two-dimensional (2D) material showing various polymorphs stable at ambient condition. Recently, a new phase with a layered hexagonal lattice (γ-GeSe) was synthesized with ambient stability and extraordinary electronic conductivity even higher than graphite while its monolayer is semiconducting. In this work, via using first-principles derived force constants and Boltzmann transport theory we explore the lattice thermal conductivity (κ_l) of the monolayer γ-GeSe, together with a comparison with monolayer α-GeSe and β-GeSe. The κ_l of γ-phase is relatively low (5.50 W/mK), comparable with those of α- and β- phases. The acoustic branches in α-GeSe are well separated from the optical branches, limiting scattering channels in the phase space, while for \b{eta}-GeSe and γ-GeSe the acoustic branches are resonant with the low-frequency optical branches facilitating more phonon-phonon scattering. For γ-GeSe, the cumulative κ_l is isotropic and phononic representative mean free path (rMFP) is the shortest (17.07 nm) amongst the three polymorphs, indicating that the κ_l of the γ phase is less likely to be affected by the size of the sample, while for α-GeSe the cumulative κ_l grows slowly with mean free path and the rMFP is longer (up to 20.56 and 35.94 nm along zigzag and armchair direction, respectively), showing a stronger size-dependence of κ_l. Our work suggests that GeSe polymorphs with overall low thermal conductivity are promising contenders for thermoelectric and thermal management applications.

cond-mat.mtrl-sci↗

arXiv2022

Multi-Task Mixture Density Graph Neural Networks for Predicting Cu-based Single-Atom Alloy Catalysts for CO2 Reduction Reaction

Graph neural networks (GNNs) have drawn more and more attention from material scientists and demonstrated a high capacity to establish connections between the structure and properties. However, with only unrelaxed structures provided as input, few GNN models can predict the thermodynamic properties of relaxed configurations with an acceptable level of error. In this work, we develop a multi-task (MT) architecture based on DimeNet++ and mixture density networks to improve the performance of such task. Taking CO adsorption on Cu-based single-atom alloy catalysts as an illustration, we show that our method can reliably estimate CO adsorption energy with a mean absolute error of 0.087 eV from the initial CO adsorption structures without costly first-principles calculations. Further, compared to other state-of-the-art GNN methods, our model exhibits improved generalization ability when predicting catalytic performance of out-of-domain configurations, built with either unseen substrate surfaces or doping species. We show that the proposed MT GNN strategy can facilitate catalyst discovery.

cond-mat.mtrl-sci↗

arXiv2022

Manifold Optimization Methods for Hybrid beamforming in mmWave Dual-Function Radar-Communication System

As a cost-effective alternative, hybrid analog and digital beamforming architecture is a promising scheme for millimeter wave (mmWave) system. This paper considers two hybrid beamforming architectures, i.e. the partially-connected and fully-connected structures, for mmWave dual-function radar communication (DFRC) system, where the transmitter communicates with the downlink users and detects radar targets simultaneously. The optimization problems are formulated by minimizing a weighted summation of radar and communication performance, subject to constant modulus and power constraints. To tackle the non-convexities caused by the two resultant problems, effective Riemannian optimization algorithms are proposed. Specifically, for the fully-connected structure, a manifold algorithm based on the alternating direction method of multipliers (ADMM) is developed. While for the partially-connected structure, a low-complexity Riemannian product manifold trust region (RPM-TR) algorithm is proposed to approach the near-optional solution. Numerical simulations are provided to demonstrate the effectiveness of the proposed methods.

eess.SP↗

arXiv2022

Fast Intercalation of Lithium in Semi-Metallic γ-GeSe Nanosheet: A New Group-IV Monochalcogenide for Lithium-Ion Battery Application

Existence of van der Waals gaps renders two-dimensional (2D) materials ideal passages of lithium for being used as anode materials. However, the requirement of good conductivity significantly limits the choice of 2D candidates. So far only graphite is satisfying due to its relatively high conductivity. Recently, a new polymorph of layered germanium selenide (Gamma-GeSe) was proven to be semimetal in its bulk phase with a higher conductivity than graphite while its monolayer behaves semiconducting. In this work, by using first-principles calculations, we examined the possibility of using this new group-IV monochalcogenide, Gamma-GeSe, as anode in the Li-ion battery (LIBs). Our studies revealed that Li atom would form an ionic adsorption with adjacent selenium atoms at the hollow site and exist in cationic state (lost 0.89 e to Gamma-GeSe). Results of climbing image-nudged elastic band show the diffusion barrier of Li is 0.21 eV in the monolayer limit, which can activate a relatively fast diffusion even at room temperature on the Gamma-GeSe surface. The calculated theoretical average voltages range from 0.071 to 0.015 V at different stoichiometry of LixGeSe with minor volume variation, suggesting its potential application as anode of LIBs. The predicted moderate binding energy, a low open circuit voltage (comparable to graphite) and a fast motion of Li suggests that Gamma-GeSe nanosheet can be chemically exfoliated via Li intercalation and a promising candidate as the anode material for LIBs.

cond-mat.mtrl-sci↗

arXiv2022

Size and Stoichiometric Dependence of Thermal Conductivities of InxGa1-xN: A Molecular Dynamics Study

The thermal conductivities k of wurtzite InxGa1-xN are investigated using equilibrium molecular dynamics (MD) method. The k of InxGa1-xN rapidly declines from InN (k_InN = 141 W/mK) or GaN (k_GaN = 500 W/mK) to InxGa1-xN, and reaches a minimum (k_min = 19 W/mK) when x is around 0.5 at 300 K. The mean free path (MFP) of InxGa1-xN, ranging from 2 to 5 nm and following the same trend with the k, is extrapolated in our simulation and a parabolic relationship between x and MFP is established. We find that the k of InxGa1-xN decreases with increasing temperatures. The evolution of k of InxGa1-xN is also examined by projecting the momentum-energy relationship of phonons from MD trajectories. The phonon dispersion and phonon density of states for InxGa1-xN reflect a slightly more flattened dispersive phononic curve of the alloying system. Despite an overestimated k than experimental values, our calculated k at 300 K agrees well with the results obtained by solving Boltzmann transport equation and also has the same stoichiometric trend with the experimental data. Our study provides the coherent analysis of the effect of thickness, temperature and stoichiometric content on the thermal transport of InxGa1-xN which is helpful for the thermal management of InxGa1-xN based devices.

cond-mat.mtrl-sci↗

arXiv2022

Efficient Passivation of Surface Defects by Lewis Base in Lead-free Tin-based Perovskite Solar Cells

Lead-free tin-based perovskites are highly appealing for the next generation of solar cells due to their intriguing optoelectronic properties. However, the tendency of Sn2+ oxidation to Sn4+ in the tin-based perovskites induces serious film degradation and performance deterioration. Herein, we demonstrate, through the density functional theory based first-principle calculations in a surface slab model, that the surface defects of the Sn-based perovskite FASnI3 (FA = NH2CHNH2+) could be effectively passivated by the Lewis base molecules. The passivation performance of Lewis base molecules in tin-based perovskite is tightly correlated with their molecular hardness. We reveal that the degree of hardness of Lewis adsorbate governs the stabilization via dual effects: first, changing the stubborn spatial distribution of tin vacancy (VSn) by triggering charge redistribution; second, saturating the dangling states while simultaneously reducing the amounts of deep band gap states. Specifically, the hard Lewis base molecules like edamine (N-donor group) and Isatin-Cl (Cl-donor group) would show a better healing effect than other candidates on the defects-contained tin-based perovskite surface with a somehow hard Lewis acid nature. Our research provides a general strategy for additive engineering and fabricating stable and high-efficiency lead-free Sn-based perovskite solar cells.

physics.chem-ph↗

arXiv2022

Strong Reduction of Thermal Conductivity of WSe2 with Introduction of Atomic Defects

The thermal conductivities of pristine and defective tungsten diselenide (WSe2) are investigated by using equilibrium molecular dynamics method. The thermal conductivity of WSe2 increases dramatically with size below a characteristic with of ~ 5 nm and levels off for broader samples and reaches a constant value of ~2 W/mK. By introducing atomic vacancies, we discovered that the thermal conductivity of WSe2 is significantly reduced. In particular, the W vacancy has a greater impact on thermal conductivity reduction than Se vacancies: the thermal conductivity of pristine WSe2 reduced by ~60% and ~70% with the adding of ~1% of Se and W vacancies, respectively. The reduction of thermal conductivity is found to be related with the decrease of mean free path (MFP) of phonons in the defective WSe2. The MFP of WSe2 decreases from ~4.2 nm for prefect WSe2 to ~2.2 nm with the adding of 0.9% Se vacancies. More sophisticated types of point defects, such as vacancy clusters and anti-site defects, are explored in addition to single vacancies, and are found to dramatically renormalize the phonons. The reconstruction of the bonds leads to localized phonons in the forbidden gap in the phonon density of states which leads to the drop of thermal conduction. This work demonstrates the influence of different defects on thermal conductivity of single-layer WSe2, providing insight into the process of defect-induced phonon transport as well as ways to improve heat dissipation in WSe2-based electronic devices.

cond-mat.mtrl-sci↗

arXiv2022

LHNN: Lattice Hypergraph Neural Network for VLSI Congestion Prediction

Precise congestion prediction from a placement solution plays a crucial role in circuit placement. This work proposes the lattice hypergraph (LH-graph), a novel graph formulation for circuits, which preserves netlist data during the whole learning process, and enables the congestion information propagated geometrically and topologically. Based on the formulation, we further developed a heterogeneous graph neural network architecture LHNN, jointing the routing demand regression to support the congestion spot classification. LHNN constantly achieves more than 35% improvements compared with U-nets and Pix2Pix on the F1 score. We expect our work shall highlight essential procedures using machine learning for congestion prediction.

cs.LG↗

arXiv2021

SCOUTER: Slot Attention-based Classifier for Explainable Image Recognition

Explainable artificial intelligence has been gaining attention in the past few years. However, most existing methods are based on gradients or intermediate features, which are not directly involved in the decision-making process of the classifier. In this paper, we propose a slot attention-based classifier called SCOUTER for transparent yet accurate classification. Two major differences from other attention-based methods include: (a) SCOUTER's explanation is involved in the final confidence for each category, offering more intuitive interpretation, and (b) all the categories have their corresponding positive or negative explanation, which tells "why the image is of a certain category" or "why the image is not of a certain category." We design a new loss tailored for SCOUTER that controls the model's behavior to switch between positive and negative explanations, as well as the size of explanatory regions. Experimental results show that SCOUTER can give better visual explanations in terms of various metrics while keeping good accuracy on small and medium-sized datasets.

cs.CV↗

arXiv2021

A systematic mapping study on security countermeasures of in-vehicle communication systems

The innovations of vehicle connectivity have been increasing dramatically to enhance the safety and user experience of driving, while the rising numbers of interfaces to the external world also bring security threats to vehicles. Many security countermeasures have been proposed and discussed to protect the systems and services against attacks. To provide an overview of the current states in this research field, we conducted a systematic mapping study on the topic area "security countermeasures of in-vehicle communication systems". 279 papers are identified based on the defined study identification strategy and criteria. We discussed four research questions related to the security countermeasures, validation methods, publication patterns, and research trends and gaps based on the extracted and classified data. Finally, we evaluated the validity threats, the study identification results, and the whole mapping process. We found that the studies in this topic area are increasing rapidly in recent years. However, there are still gaps in various subtopics like automotive Ethernet security, anomaly reaction, and so on. This study reviews the target field not only related to research findings but also research activities, which can help identify research gaps at a high level and inspire new ideas for future work.

cs.CR↗

arXiv2021

The SARAS Endoscopic Surgeon Action Detection (ESAD) dataset: Challenges and methods

For an autonomous robotic system, monitoring surgeon actions and assisting the main surgeon during a procedure can be very challenging. The challenges come from the peculiar structure of the surgical scene, the greater similarity in appearance of actions performed via tools in a cavity compared to, say, human actions in unconstrained environments, as well as from the motion of the endoscopic camera. This paper presents ESAD, the first large-scale dataset designed to tackle the problem of surgeon action detection in endoscopic minimally invasive surgery. ESAD aims at contributing to increase the effectiveness and reliability of surgical assistant robots by realistically testing their awareness of the actions performed by a surgeon. The dataset provides bounding box annotation for 21 action classes on real endoscopic video frames captured during prostatectomy, and was used as the basis of a recent MIDL 2020 challenge. We also present an analysis of the dataset conducted using the baseline model which was released as part of the challenge, and a description of the top performing models submitted to the challenge together with the results they obtained. This study provides significant insight into what approaches can be effective and can be extended further. We believe that ESAD will serve in the future as a useful benchmark for all researchers active in surgeon action detection and assistive robotics at large.

cs.CV↗

arXiv2020

Match Them Up: Visually Explainable Few-shot Image Classification

Few-shot learning (FSL) approaches are usually based on an assumption that the pre-trained knowledge can be obtained from base (seen) categories and can be well transferred to novel (unseen) categories. However, there is no guarantee, especially for the latter part. This issue leads to the unknown nature of the inference process in most FSL methods, which hampers its application in some risk-sensitive areas. In this paper, we reveal a new way to perform FSL for image classification, using visual representations from the backbone model and weights generated by a newly-emerged explainable classifier. The weighted representations only include a minimum number of distinguishable features and the visualized weights can serve as an informative hint for the FSL process. Finally, a discriminator will compare the representations of each pair of the images in the support set and the query set. Pairs with the highest scores will decide the classification results. Experimental results prove that the proposed method can achieve both good accuracy and satisfactory explainability on three mainstream datasets.

cs.CV↗

arXiv2020

Noisy-LSTM: Improving Temporal Awareness for Video Semantic Segmentation

Semantic video segmentation is a key challenge for various applications. This paper presents a new model named Noisy-LSTM, which is trainable in an end-to-end manner, with convolutional LSTMs (ConvLSTMs) to leverage the temporal coherency in video frames. We also present a simple yet effective training strategy, which replaces a frame in video sequence with noises. This strategy spoils the temporal coherency in video frames during training and thus makes the temporal links in ConvLSTMs unreliable, which may consequently improve feature extraction from video frames, as well as serve as a regularizer to avoid overfitting, without requiring extra data annotation or computational costs. Experimental results demonstrate that the proposed model can achieve state-of-the-art performances in both the CityScapes and EndoVis2018 datasets.

cs.CV↗

arXiv2020

Intrinsic Transverse Momentum and Evolution in Weighted Spin Asymmetries

The transverse momentum dependent (TMD) and collinear higher twist theoretical factorization frameworks are the most frequently used approaches to describing spin dependent hard cross sections weighted by and integrated over transverse momentum. Of particular interest is the contribution from small transverse momentum associated with the target bound state. In phenomenological applications, this contribution is often investigated using transverse momentum weighted integrals that sharply regulate the large transverse momentum contribution, for example with Gaussian parametrizations. Since the result is a kind of hybrid of TMD and collinear (inclusive) treatments, it is important to establish if and how the formalisms are related in applications to weighted integral observables. The suppression of a large transverse momentum tail, for example, can potentially affect the type of evolution that is applicable. We find that a naive version of a widely used identity relating the $k_T^2$-weighted and integrated Sivers TMD function to a renormalized twist-3 function has strongly ambiguous ultraviolet contributions, and that corrections to it are not necessarily perturbatively suppressed. We discuss the implications for applications, arguing in particular that the relevant evolution for transverse momentum weighted and integrated cross sections with sharp effective large transverse momentum cutoffs is of the TMD form rather than the standard renormalization group evolution of collinear correlation functions.

hep-ph↗