arXiv subjects

Xin Du

Publications and source records attributed to Xin Du.

At least 73 records · Page 4Linked to original sources

arXiv2022

FreSCo: Frequency-Domain Scan Context for LiDAR-based Place Recognition with Translation and Rotation Invariance

Place recognition plays a crucial role in re-localization and loop closure detection tasks for robots and vehicles. This paper seeks a well-defined global descriptor for LiDAR-based place recognition. Compared to local descriptors, global descriptors show remarkable performance in urban road scenes but are usually viewpoint-dependent. To this end, we propose a simple yet robust global descriptor dubbed FreSCo that decomposes the viewpoint difference during revisit and achieves both translation and rotation invariance by leveraging Fourier Transform and circular shift technique. Besides, a fast two-stage pose estimation method is proposed to estimate the relative pose after place retrieval by utilizing the compact 2D point clouds extracted from the original data. Experiments show that FreSCo exhibited superior performance than contemporaneous methods on sequences of different scenes from multiple datasets. Code will be publicly available at https://github.com/soytony/FreSCo.

cs.RO↗

arXiv2022

Rethinking Performance Gains in Image Dehazing Networks

Image dehazing is an active topic in low-level vision, and many image dehazing networks have been proposed with the rapid development of deep learning. Although these networks' pipelines work fine, the key mechanism to improving image dehazing performance remains unclear. For this reason, we do not target to propose a dehazing network with fancy modules; rather, we make minimal modifications to popular U-Net to obtain a compact dehazing network. Specifically, we swap out the convolutional blocks in U-Net for residual blocks with the gating mechanism, fuse the feature maps of main paths and skip connections using the selective kernel, and call the resulting U-Net variant gUNet. As a result, with a significantly reduced overhead, gUNet is superior to state-of-the-art methods on multiple image dehazing datasets. Finally, we verify these key designs to the performance gain of image dehazing networks through extensive ablation studies.

cs.CV↗

arXiv2022

Modular Degradation Simulation and Restoration for Under-Display Camera

Under-display camera (UDC) provides an elegant solution for full-screen smartphones. However, UDC captured images suffer from severe degradation since sensors lie under the display. Although this issue can be tackled by image restoration networks, these networks require large-scale image pairs for training. To this end, we propose a modular network dubbed MPGNet trained using the generative adversarial network (GAN) framework for simulating UDC imaging. Specifically, we note that the UDC imaging degradation process contains brightness attenuation, blurring, and noise corruption. Thus we model each degradation with a characteristic-related modular network, and all modular networks are cascaded to form the generator. Together with a pixel-wise discriminator and supervised loss, we can train the generator to simulate the UDC imaging degradation process. Furthermore, we present a Transformer-style network named DWFormer for UDC image restoration. For practical purposes, we use depth-wise convolution instead of the multi-head self-attention to aggregate local spatial information. Moreover, we propose a novel channel attention module to aggregate global information, which is critical for brightness recovery. We conduct evaluations on the UDC benchmark, and our method surpasses the previous state-of-the-art models by 1.23 dB on the P-OLED track and 0.71 dB on the T-OLED track, respectively.

eess.IV↗

arXiv2022

Multi-Curve Translator for High-Resolution Photorealistic Image Translation

The dominant image-to-image translation methods are based on fully convolutional networks, which extract and translate an image's features and then reconstruct the image. However, they have unacceptable computational costs when working with high-resolution images. To this end, we present the Multi-Curve Translator (MCT), which not only predicts the translated pixels for the corresponding input pixels but also for their neighboring pixels. And if a high-resolution image is downsampled to its low-resolution version, the lost pixels are the remaining pixels' neighboring pixels. So MCT makes it possible to feed the network only the downsampled image to perform the mapping for the full-resolution image, which can dramatically lower the computational cost. Besides, MCT is a plug-in approach that utilizes existing base models and requires only replacing their output layers. Experiments demonstrate that the MCT variants can process 4K images in real-time and achieve comparable or even better performance than the base models on various photorealistic image-to-image translation tasks.

cs.CV↗

arXiv2022

A Low-latency Communication Design for Brain Simulations

Brain simulation, as one of the latest advances in artificial intelligence, facilitates better understanding about how information is represented and processed in the brain. The extreme complexity of human brain makes brain simulations only feasible upon high-performance computing platforms. Supercomputers with a large number of interconnected graphical processing units (GPUs) are currently employed for supporting brain simulations. Therefore, high-throughput low-latency inter-GPU communications in supercomputers play a crucial role in meeting the performance requirements of brain simulation as a highly time-sensitive application. In this paper, we first provide an overview of the current parallelizing technologies for brain simulations using multi-GPU architectures. Then, we analyze the challenges to communications for brain simulation and summarize guidelines for communication design to address such challenges. Furthermore, we propose a partitioning algorithm and a two-level routing method to achieve efficient low-latency communications in multi-GPU architecture for brain simulation. We report experiment results obtained on a supercomputer with 2,000 GPUs for simulating a brain model with 10 billion neurons to show that our approach can significantly improve communication performance. We also discuss open issues and identify some research directions for low-latency communication design for brain simulations.

cs.DC↗

arXiv2022

Scientific Workflows in Heterogeneous Edge-Cloud Computing: A Data Placement Strategy Based on Reinforcement learning

The heterogeneous edge-cloud computing paradigm can provide an optimal solution to deploy scientific workflows compared to cloud computing or other traditional distributed computing environments. Owing to the different sizes of scientific datasets and the privacy issue concerning some of these datasets, it is essential to find a data placement strategy that can minimize data transmission time. Some state-of-the-art data placement strategies combine edge computing and cloud computing to distribute scientific datasets. However, the dynamic distribution of newly generated datasets to appropriate datacenters and exiting the spent datasets are still a challenge during workflows execution. To address this challenge, this study not only constructs a data placement model that includes shared datasets within individual and among multiple workflows across various geographical regions, but also proposes a data placement strategy (DYM-RL-DPS) based on algorithms of two stages. First, during the build-time stage of workflows, we use the discrete particle swarm optimization algorithm with differential evolution to pre-allocate initial datasets to proper datacenters. Then, we reformulate the dynamic datasets distribution problem as a Markov decision process and provide a reinforcement learning-based approach to learn the optimal strategy in the runtime stage of scientific workflows. Through simulating heterogeneous edge-cloud computing environments, we designed comprehensive experiments to demonstrate the superiority of DYM-RL-DPS. The results of our strategy can effectively reduce the data transmission time as compared to other strategies.

cs.DC↗

arXiv2022

NLOS Error Mitigation Using Weighted Least Squares and Kalman Filter in UWB Positioning

In wireless positioning systems, non-line-of-sight (NLOS) is a challenging problem. NLOS causes great ranging bias and location error, so NLOS mitigation is essential for high accuracy positioning. In this letter, we propose the Weighted-Least-Squares Robust Kalman Filter (WLS-RKF) for NLOS identification and mitigation. WLS-RKF employs a hypothesis test based on Mahalanobis distance for NLOS identification, and updates the corresponding Kalman filter using the WLS solution. It requires no prior knowledge about NLOS distribution or signal features. We perform simulations and experiments for ultra-wideband (UWB) positioning in various scenarios. The results confirm that WLS-RKF effectively mitigates NLOS error and achieves 5cm positioning accuracy.

eess.SY↗

arXiv2022

Vision Transformers for Single Image Dehazing

Image dehazing is a representative low-level vision task that estimates latent haze-free images from hazy images. In recent years, convolutional neural network-based methods have dominated image dehazing. However, vision Transformers, which has recently made a breakthrough in high-level vision tasks, has not brought new dimensions to image dehazing. We start with the popular Swin Transformer and find that several of its key designs are unsuitable for image dehazing. To this end, we propose DehazeFormer, which consists of various improvements, such as the modified normalization layer, activation function, and spatial information aggregation scheme. We train multiple variants of DehazeFormer on various datasets to demonstrate its effectiveness. Specifically, on the most frequently used SOTS indoor set, our small model outperforms FFA-Net with only 25% #Param and 5% computational cost. To the best of our knowledge, our large model is the first method with the PSNR over 40 dB on the SOTS indoor set, dramatically outperforming the previous state-of-the-art methods. We also collect a large-scale realistic remote sensing dehazing dataset for evaluating the method's capability to remove highly non-homogeneous haze.

cs.CV↗

arXiv2022

Vision Checklist: Towards Testable Error Analysis of Image Models to Help System Designers Interrogate Model Capabilities

Using large pre-trained models for image recognition tasks is becoming increasingly common owing to the well acknowledged success of recent models like vision transformers and other CNN-based models like VGG and Resnet. The high accuracy of these models on benchmark tasks has translated into their practical use across many domains including safety-critical applications like autonomous driving and medical diagnostics. Despite their widespread use, image models have been shown to be fragile to changes in the operating environment, bringing their robustness into question. There is an urgent need for methods that systematically characterise and quantify the capabilities of these models to help designers understand and provide guarantees about their safety and robustness. In this paper, we propose Vision Checklist, a framework aimed at interrogating the capabilities of a model in order to produce a report that can be used by a system designer for robustness evaluations. This framework proposes a set of perturbation operations that can be applied on the underlying data to generate test samples of different types. The perturbations reflect potential changes in operating environments, and interrogate various properties ranging from the strictly quantitative to more qualitative. Our framework is evaluated on multiple datasets like Tinyimagenet, CIFAR10, CIFAR100 and Camelyon17 and for models like ViT and Resnet. Our Vision Checklist proposes a specific set of evaluations that can be integrated into the previously proposed concept of a model card. Robustness evaluations like our checklist will be crucial in future safety evaluations of visual perception modules, and be useful for a wide range of stakeholders including designers, deployers, and regulators involved in the certification of these systems. Source code of Vision Checklist would be open for public use.

cs.CV↗

arXiv2021

V2V-Based Task Offloading and Resource Allocation in Vehicular Edge Computing Networks

In the research and application of vehicle ad hoc networks (VANETs), it is often assumed that vehicles obtain cloud computing services by accessing to roadside units (RSUs). However, due to the problems of insufficient construction quantity, limited communication range and overload of calculation load of roadside units, the calculation mode relying only on vehicle to roadside units is difficult to deal with complex and changeable calculation tasks. In this paper, when the roadside unit is missing, the vehicle mobile unit is regarded as a natural edge computing node to make full use of the excess computing power of mobile vehicles and perform the offloading task of surrounding mobile vehicles in time. In this paper, the OPFTO framework is designed, an improved task allocation algorithm HGSA is proposed, and the pre-filtering process is designed with full consideration of the moving characteristics of vehicles. In addition, vehicle simulation experiments show that the proposed strategy has the advantages of low delay and high accuracy compared with other task scheduling strategies, which provides a reference scheme for the construction of Urban Intelligent Transportation in the future.

cs.DC↗

arXiv2021

DPVI: A Dynamic-Weight Particle-Based Variational Inference Framework

The recently developed Particle-based Variational Inference (ParVI) methods drive the empirical distribution of a set of \emph{fixed-weight} particles towards a given target distribution $π$ by iteratively updating particles' positions. However, the fixed weight restriction greatly confines the empirical distribution's approximation ability, especially when the particle number is limited. In this paper, we propose to dynamically adjust particles' weights according to a Fisher-Rao reaction flow. We develop a general Dynamic-weight Particle-based Variational Inference (DPVI) framework according to a novel continuous composite flow, which evolves the positions and weights of particles simultaneously. We show that the mean-field limit of our composite flow is actually a Wasserstein-Fisher-Rao gradient flow of certain dissimilarity functional $\mathcal{F}$, which leads to a faster decrease of $\mathcal{F}$ than the Wasserstein gradient flow underlying existing fixed-weight ParVIs. By using different finite-particle approximations in our general framework, we derive several efficient DPVI algorithms. The empirical results demonstrate the superiority of our derived DPVI algorithms over their fixed-weight counterparts.

cs.LG↗

arXiv2021

Multiple magnetism controlled topological states in EuAgAs

The interplay between magnetism and band topology is a focus of current research on magnetic topological systems. Based on first-principle calculations and symmetry analysis, we reveal multiple intriguing topological states can be realized in a single system EuAgAs, controlled by the magnetic ordering. The material is Dirac semimetal in the paramagnetic state, with a pair of accidental Dirac points. Under different magnetic configurations, the Dirac points can evolve into magnetic triply-degenerate points, magnetic linear and double Weyl points, or being gapped out and making the system a topological mirror semimetal characterized by mirror Chern numbers. The change in bulk topology is also manifested in the surface states, including the surface Fermi arcs and surface Dirac cones. In addition, the antiferromagnetic states also feature a nontrivial Z4 index, implying a higher order topology. These results deepen our understanding of magnetic topological states and provide new perspectives for spintronic applications.

cond-mat.mtrl-sci↗

arXiv2021

An unconstrained-like control-based dynamic method for optimization problems with simple bounds

The optimization problems with simple bounds are an important class of problems. To facilitate the computation of such problems, an unconstrained-like dynamic method, motivated by the Lyapunov control principle, is proposed. This method employs the anti-windup limited integrator to address the bounds of parameters upon the dynamics for unconstrained problem, and then solves the transformed Initial-value Problems (IVPs) with mature Ordinary Differential Equation (ODE) integration methods. It is proved that when the gain matrix is diagonal, the result is equivalent to that of the general dynamic method which involves an intermediate Quadratic Programming (QP) sub-problem. Thus, the global convergence to the optimal solution is guaranteed without the requirement of the strict complementarity condition. Since the estimation of the right active constraints is avoided and no programming sub-problem is involved in the computation process, it shows higher efficiency than the general dynamic method and other common iterative methods through the numerical examples. In particular, the implementation is simple, and the proposed method is easy-to-use.

math.OC↗

arXiv2021

Adaptive Frequency-limited H2-Model Order Reduction

In this paper, we present an adaptive framework for constructing a pseudo-optimal reduced model for the frequency-limited H2-optimal model order reduction problem. We show that the frequency-limited pseudo-optimal reduced-order model has an inherent property of monotonic decay in error if the interpolation points and tangential directions are selected appropriately. We also show that this property can be used to make an automatic selection of the order of the reduced model for an allowable tolerance in error. The proposed algorithm adaptively increases the order of the reduced model such that the frequency-limited H2-norm error decays monotonically irrespective of the choice of interpolation points and tangential directions. The stability of the reduced-order model is also guaranteed. Additionally, it also generates the approximations of the frequency-limited system Gramians that monotonically approach the original solution. Further, we show that the low-rank alternating direction implicit iteration method for solving large-scale frequency-limited Lyapunov equations implicitly performs frequency-limited pseudo-optimal model order reduction. We consider two numerical examples to validate the theory presented in the paper.

eess.SY↗

arXiv2021

On frequency- and time-limited H2-optimal model order reduction

In this paper, the problems of frequency-limited and time-limited H2-optimal model order reduction of linear time-invariant systems are considered within the oblique projection framework. It is shown that it is inherently not possible to satisfy all the necessary conditions for the local minimizer in the oblique projection framework. The conditions for exact satisfaction of the optimality conditions are also discussed. Further, the equivalence between the tangential interpolation conditions and the gramians-based necessary condition for the local optimum is established. Based on this equivalence, iterative algorithms that nearly satisfy these interpolation-based necessary conditions are proposed. The deviation in satisfaction of the optimality conditions decay as the order of the reduced-model is increased in the proposed algorithms. Moreover, stationary point iteration algorithms that satisfy two out of three necessary conditions for the local minimizer are also proposed. There also, the deviation in satisfaction of the third optimality conditions decay as the order of the reduced-model is increased in the proposed algorithms. The efficacy of the proposed algorithms is validated by considering one illustrative and three high-order models that are considered a benchmark for testing model order reduction algorithms.

eess.SY↗

arXiv2021

StarEnhancer: Learning Real-Time and Style-Aware Image Enhancement

Image enhancement is a subjective process whose targets vary with user preferences. In this paper, we propose a deep learning-based image enhancement method covering multiple tonal styles using only a single model dubbed StarEnhancer. It can transform an image from one tonal style to another, even if that style is unseen. With a simple one-time setting, users can customize the model to make the enhanced images more in line with their aesthetics. To make the method more practical, we propose a well-designed enhancer that can process a 4K-resolution image over 200 FPS but surpasses the contemporaneous single style image enhancement methods in terms of PSNR, SSIM, and LPIPS. Finally, our proposed enhancement method has good interactability, which allows the user to fine-tune the enhanced image using intuitive options.

cs.CV↗

arXiv2021

Edge service resource allocation strategy based on intelligent prediction

Artificial intelligence is one of the important technologies for industrial applications, but it requires a lot of computing resources and sensor data to support it. With the development of edge computing and the Internet of Things, artificial intelligence are playing an increasingly important role in the field of edge services. Therefore, how to make intelligent algorithms provide better services and the development of the Internet of Things has become an increasingly important topic. This paper focuses on the application of edge service distribution strategy, and proposes an edge service distribution strategy based on intelligent prediction, which reduces the bandwidth consumption of edge service providers and minimizes the cost of edge service providers. In addition, this article uses the real data provided by the Wangsu Technology Company and an improved long and short term memory prediction method to dynamically change the bandwidth, and achieves better optimization of resources allocation comparing with actual industrial applications.The simulation results show that our intelligent prediction can achieve good results, and the mechanism can achieve higher resource utilization.

cs.DC↗

arXiv2021

Moire Superlattice Modulations in Single-Unit-Cell FeTe Films Grown on NbSe2 Single Crystals

Interface can be a fertile ground for exotic quantum states, including topological superconductivity, Majorana mode, fractal quantum Hall effect, unconventional superconductivity, Mott insulator, etc. Here we grow single-unit-cell (1UC) FeTe film on NbSe2 single crystal by molecular beam epitaxy (MBE) and investigate the film in-situ with home-made cryogenic scanning tunneling microscopy (STM) and non-contact atomic force microscopy (AFM) combined system. We find different stripe-like superlattice modulations on grown FeTe film with different misorientation angles with respect to NbSe2 substrate. We show that these stripe-like superlattice modulations can be understood as moire pattern forming between FeTe film and NbSe2 substrate. Our results indicate that the interface between FeTe and NbSe2 is atomically sharp. By STM-AFM combined measurement, we suggest the moire superlattice modulations have an electronic origin when the misorientation angle is relatively small (<= 3 degree) and have structural relaxation when the misorientation angle is relatively large (>= 10 degree).

cond-mat.str-el↗