arXiv subjects

Wei Sun

Publications and source records attributed to Wei Sun.

At least 289 records · Page 16Linked to original sources

arXiv2023

VDPVE: VQA Dataset for Perceptual Video Enhancement

Recently, many video enhancement methods have been proposed to improve video quality from different aspects such as color, brightness, contrast, and stability. Therefore, how to evaluate the quality of the enhanced video in a way consistent with human visual perception is an important research topic. However, most video quality assessment methods mainly calculate video quality by estimating the distortion degrees of videos from an overall perspective. Few researchers have specifically proposed a video quality assessment method for video enhancement, and there is also no comprehensive video quality assessment dataset available in public. Therefore, we construct a Video quality assessment dataset for Perceptual Video Enhancement (VDPVE) in this paper. The VDPVE has 1211 videos with different enhancements, which can be divided into three sub-datasets: the first sub-dataset has 600 videos with color, brightness, and contrast enhancements; the second sub-dataset has 310 videos with deblurring; and the third sub-dataset has 301 deshaked videos. We invited 21 subjects (20 valid subjects) to rate all enhanced videos in the VDPVE. After normalizing and averaging the subjective opinion scores, the mean opinion score of each video can be obtained. Furthermore, we split the VDPVE into a training set, a validation set, and a test set, and verify the performance of several state-of-the-art video quality assessment methods on the test set of the VDPVE.

eess.IV↗

arXiv2023

Learning Audio-Visual Dereverberation

Reverberation not only degrades the quality of speech for human perception, but also severely impacts the accuracy of automatic speech recognition. Prior work attempts to remove reverberation based on the audio modality only. Our idea is to learn to dereverberate speech from audio-visual observations. The visual environment surrounding a human speaker reveals important cues about the room geometry, materials, and speaker location, all of which influence the precise reverberation effects. We introduce Visually-Informed Dereverberation of Audio (VIDA), an end-to-end approach that learns to remove reverberation based on both the observed monaural sound and visual scene. In support of this new task, we develop a large-scale dataset SoundSpaces-Speech that uses realistic acoustic renderings of speech in real-world 3D scans of homes offering a variety of room acoustics. Demonstrating our approach on both simulated and real imagery for speech enhancement, speech recognition, and speaker identification, we show it achieves state-of-the-art performance and substantially improves over audio-only methods.

cs.SD↗

arXiv2023

Foundry manufacturing of tight-confinement, dispersion-engineered, ultralow-loss silicon nitride photonic integrated circuit

The foundry development of integrated photonics has revolutionized today's optical interconnect and datacenters. Over the last decade, we have witnessed the rising of silicon nitride (Si$_3$N$_4$) integrated photonics, which is currently transferring from laboratory research to foundry manufacturing. The development and transition are triggered by the ultimate need of low optical loss offered by Si$_3$N$_4$, which is beyond the reach of silicon and III-V semiconductors. Combined with modest Kerr nonlinearity, tight optical confinement and dispersion engineering, Si$_3$N$_4$ has today become the leading platform for linear and Kerr nonlinear photonics, and has enabled chip-scale lasers featuring ultralow noise on par with table-top fiber lasers. However, so far all the reported fabrication processes of tight-confinement, dispersion-engineered Si$_3$N$_4$ photonic integrated circuit (PIC) with optical loss down to few dB/m have only been developed on 4-inch or smaller wafers. Yet, to transfer these processes to established CMOS foundries that typically operate 6-inch or even larger wafers, challenges remain. In this work, we demonstrate the first foundry-standard fabrication process of Si$_3$N$_4$ PIC with only 2.6 dB/m loss, thickness above 800 nm, and near 100% fabrication yield on 6-inch wafers. Such thick and ultralow-loss Si$_3$N$_4$ PIC enables low-threshold generation of soliton frequency combs. Merging with advanced heterogeneous integration, active ultralow-loss Si$_3$N$_4$ integrated photonics could pave an avenue to addressing future demands in our increasingly information-driven society.

physics.optics↗

arXiv2023

Full Point Encoding for Local Feature Aggregation in 3D Point Clouds

Point cloud processing methods exploit local point features and global context through aggregation which does not explicity model the internal correlations between local and global features. To address this problem, we propose full point encoding which is applicable to convolution and transformer architectures. Specifically, we propose Full Point Convolution (FPConv) and Full Point Transformer (FPTransformer) architectures. The key idea is to adaptively learn the weights from local and global geometric connections, where the connections are established through local and global correlation functions respectively. FPConv and FPTransformer simultaneously model the local and global geometric relationships as well as their internal correlations, demonstrating strong generalization ability and high performance. FPConv is incorporated in classical hierarchical network architectures to achieve local and global shape-aware learning. In FPTransformer, we introduce full point position encoding in self-attention, that hierarchically encodes each point position in the global and local receptive field. We also propose a shape aware downsampling block which takes into account the local shape and the global context. Experimental comparison to existing methods on benchmark datasets show the efficacy of FPConv and FPTransformer for semantic segmentation, object detection, classification, and normal estimation tasks. In particular, we achieve state-of-the-art semantic segmentation results of 76% mIoU on S3DIS 6-fold and 72.2% on S3DIS Area5.

cs.CV↗

arXiv2023

DANet: Density Adaptive Convolutional Network with Interactive Attention for 3D Point Clouds

Local features and contextual dependencies are crucial for 3D point cloud analysis. Many works have been devoted to designing better local convolutional kernels that exploit the contextual dependencies. However, current point convolutions lack robustness to varying point cloud density. Moreover, contextual modeling is dominated by non-local or self-attention models which are computationally expensive. To solve these problems, we propose density adaptive convolution, coined DAConv. The key idea is to adaptively learn the convolutional weights from geometric connections obtained from the point density and position. To extract precise context dependencies with fewer computations, we propose an interactive attention module (IAM) that embeds spatial information into channel attention along different spatial directions. DAConv and IAM are integrated in a hierarchical network architecture to achieve local density and contextual direction-aware learning for point cloud analysis. Experiments show that DAConv is significantly more robust to point density compared to existing methods and extensive comparisons on challenging 3D point cloud datasets show that our network achieves state-of-the-art classification results of 93.6% on ModelNet40, competitive semantic segmentation results of 68.71% mIoU on S3DIS and part segmentation results of 86.7% mIoU on ShapeNet.

cs.CV↗

arXiv2023

Perceptual Quality Assessment for Digital Human Heads

Digital humans are attracting more and more research interest during the last decade, the generation, representation, rendering, and animation of which have been put into large amounts of effort. However, the quality assessment of digital humans has fallen behind. Therefore, to tackle the challenge of digital human quality assessment issues, we propose the first large-scale quality assessment database for three-dimensional (3D) scanned digital human heads (DHHs). The constructed database consists of 55 reference DHHs and 1,540 distorted DHHs along with the subjective perceptual ratings. Then, a simple yet effective full-reference (FR) projection-based method is proposed to evaluate the visual quality of DHHs. The pretrained Swin Transformer tiny is employed for hierarchical feature extraction and the multi-head attention module is utilized for feature fusion. The experimental results reveal that the proposed method exhibits state-of-the-art performance among the mainstream FR metrics. The database is released at https://github.com/zzc-1998/DHHQA.

cs.CV↗

arXiv2023

Hot Brownian Motion of thermoresponsive microgels in optical tweezers shows discontinuous volume phase transition and bistability

Microgels are soft microparticles that often exhibit thermoresponsiveness and feature a transformation at a critical temperature, referred to as the volume phase transition temperature. The question of whether this transformation occurs as a smooth or as a discontinuous one is still a matter of debate. This question can be addressed by studying individual microgels trapped in optical tweezers. For this aim, composite particles were obtained by decorating pNIPAM microgels with iron oxide nanocubes. These composites become self-heating when illuminated by the infrared trapping laser, featuring Hot Brownian Motion within the trap. Above a certain laser power, a single decorated microgel features a volume phase transition that is discontinuous, while the usual continuous sigmoidal-like dependence is recovered after averaging over different microgels. The collective sigmoidal behavior enables the application of a power-to-temperature calibration and provides the effective drag coefficient of the self-heating microgels, thus establishing these composite particles as potential micro-thermometers and micro-heaters. Moreover, the self-heating microgels also exhibit an unexpected and intriguing bistability behavior above the critical temperature, probably due to partial collapses of the microgel. These results set the stage for further studies and the development of applications based on the Hot Brownian Motion of soft particles.

cond-mat.soft↗

arXiv2023

Loss Functions for Discrete Contextual Pricing with Observational Data

We study a pricing setting where each customer is offered a contextualized price based on customer and/or product features. Often only historical sales data are available, so we observe whether a customer purchased a product at the price prescribed rather than the customer's true valuation. Such observational data are influenced by historical pricing policies, which introduce difficulties in evaluating the effectiveness of future policies. The goal of this paper is to formulate loss functions that can be used for evaluating pricing policies directly from observational data, rather than going through an intermediate demand estimation stage, which may suffer from bias. To achieve this, we adapt ideas from machine learning with corrupted labels, where we consider each observed purchase decision as a known probabilistic transformation of the customer's valuation. From this transformation, we derive a class of unbiased loss functions. Within this class, we identify minimum variance estimators and estimators robust to poor demand estimation. Furthermore, we show that for contextual pricing, estimators popular in the off-policy evaluation literature fall within this class of loss functions. We offer managerial insights into scenarios under which these estimators are effective.

cs.LG↗

arXiv2023

Incipient Fault Detection in Power Distribution System: A Time-Frequency Embedded Deep Learning Based Approach

Incipient fault detection in power distribution systems is crucial to improve the reliability of the grid. However, the non-stationary nature and the inadequacy of the training dataset due to the self-recovery of the incipient fault signal, make the incipient fault detection in power distribution systems a great challenge. In this paper, we focus on incipient fault detection in power distribution systems and address the above challenges. In particular, we propose an ADaptive Time-Frequency Memory(AD-TFM) cell by embedding wavelet transform into the Long Short-Term Memory (LSTM), to extract features in time and frequency domain from the non-stationary incipient fault signals.We make scale parameters and translation parameters of wavelet transform learnable to adapt to the dynamic input signals. Based on the stacked AD-TFM cells, we design a recurrent neural network with ATtention mechanism, named AD-TFM-AT model, to detect incipient fault with multi-resolution and multi-dimension analysis. In addition, we propose two data augmentation methods, namely phase switching and temporal sliding, to effectively enlarge the training datasets. Experimental results on two open datasets show that our proposed AD-TFM-AT model and data augmentation methods achieve state-of-the-art (SOTA) performance of incipient fault detection in power distribution system. We also disclose one used dataset logged at State Grid Corporation of China to facilitate future research.

eess.SP↗

arXiv2023

Scalable Optimal Multiway-Split Decision Trees with Constraints

There has been a surge of interest in learning optimal decision trees using mixed-integer programs (MIP) in recent years, as heuristic-based methods do not guarantee optimality and find it challenging to incorporate constraints that are critical for many practical applications. However, existing MIP methods that build on an arc-based formulation do not scale well as the number of binary variables is in the order of $\mathcal{O}(2^dN)$, where $d$ and $N$ refer to the depth of the tree and the size of the dataset. Moreover, they can only handle sample-level constraints and linear metrics. In this paper, we propose a novel path-based MIP formulation where the number of decision variables is independent of $N$. We present a scalable column generation framework to solve the MIP optimally. Our framework produces a multiway-split tree which is more interpretable than the typical binary-split trees due to its shorter rules. Our method can handle nonlinear metrics such as F1 score and incorporate a broader class of constraints. We demonstrate its efficacy with extensive experiments. We present results on datasets containing up to 1,008,372 samples while existing MIP-based decision tree models do not scale well on data beyond a few thousand points. We report superior or competitive results compared to the state-of-art MIP-based methods with up to a 24X reduction in runtime.

cs.LG↗

arXiv2023

Radiative Decay Width of $J/ψ\to γη_{(2)}$ from $N_f=2$ Lattice QCD

The large radiative production rate for pseudoscalar mesons in the $J/ψ$ radiative decay remains elusive. We present the first lattice QCD calculation of partial decay width of $J/ψ$ radiatively decaying into $η_{(2)}$, the $\mathrm{SU(2)}$ flavor singlet pseudoscalar meson, which confirms QCD $\mathrm{U_A(1)}$ anomaly enhancement to the coupling of gluons with flavor singlet pseudoscalar mesons. The lattice simulation is carried out using $N_f=2$ lattice QCD gauge configurations at the pion mass $m_π \approx 350$ MeV. In particular, the distillation method has been utilized to calculate light quark loops. The results are reported here with the mass $m_{η_{(2)}}= 718(8)$ MeV and the decay width $Γ(J/ψ\toγη_{(2)})=0.385(45)$ keV. By assuming the dominance of $\mathrm{U_A(1)}$ anomaly and flavor singlet-octet mixing angle $θ=-24.5^\circ$, the production rates for the physical $η$ and $η'$ in $J/ψ$ radiative decay are predicted to be $1.15(14)\times 10^{-3}$ and $4.49(53)\times 10^{-3}$, respectively, which agree well with the experimental measurement data. Our study manifests the potential of lattice QCD studies on the light hadron production in $J/ψ$ radiative decays.

hep-lat↗

arXiv2022

X-ray Spectroscopic Evidence of Charge Exchange Emission in the Disk of M51

In the disks of spiral galaxies, diffuse soft X-ray emission is known to be strongly correlated with star-forming regions. However, this emission is not simply from a thermal-equilibrium plasma and its origin remains greatly unclear. In this work, we present an X-ray spectroscopic analysis of the emission from the northern hot spot; a region with enhanced star-formation off the nucleus of M51. Based on the high spectral resolution data from XMM-Newton/RGS observations, we unambiguously detect a high $G$ ratio ($3.2^{+6.9}_{-1.5}$) of the OVII He$α$ triplet. This high $G$ ratio is also spatially confirmed by oxygen emission-line maps from the same data. A physical model consisting of a thermal plasma and its charge exchange (CX) with neutral cool gas gives a good explanation for the $G$ ratio and the entire RGS spectra. This model also gives a satisfactory characterization of the complementary Chandra/ACIS-S data, which enables a direct imaging of the diffuse emission, tracing the hot plasma across the galaxy. The hot plasma has a similar characteristic temperature of ~0.34 keV and an approximately solar metallicity. The CX contributes ~50% to the diffuse emission in the 0.4-1.8 keV band, suggesting an effective hot/cool gas interface area about five times the geometric area of the M51 disk. Therefore, the CX appears to play a major role in the soft X-ray production and may be used as a powerful tool to probe the interface astrophysics, important for studying galactic ecosystems.

astro-ph.GA↗

arXiv2022

DQnet: Cross-Model Detail Querying for Camouflaged Object Detection

Camouflaged objects are seamlessly blended in with their surroundings, which brings a challenging detection task in computer vision. Optimizing a convolutional neural network (CNN) for camouflaged object detection (COD) tends to activate local discriminative regions while ignoring complete object extent, causing the partial activation issue which inevitably leads to missing or redundant regions of objects. In this paper, we argue that partial activation is caused by the intrinsic characteristics of CNN, where the convolution operations produce local receptive fields and experience difficulty to capture long-range feature dependency among image regions. In order to obtain feature maps that could activate full object extent, keeping the segmental results from being overwhelmed by noisy features, a novel framework termed Cross-Model Detail Querying network (DQnet) is proposed. It reasons the relations between long-range-aware representations and multi-scale local details to make the enhanced representation fully highlight the object regions and eliminate noise on non-object regions. Specifically, a vanilla ViT pretrained with self-supervised learning (SSL) is employed to model long-range dependencies among image regions. A ResNet is employed to enable learning fine-grained spatial local details in multiple scales. Then, to effectively retrieve object-related details, a Relation-Based Querying (RBQ) module is proposed to explore window-based interactions between the global representations and the multi-scale local details. Extensive experiments are conducted on the widely used COD datasets and show that our DQnet outperforms the current state-of-the-arts.

cs.CV↗

arXiv2022

Evaluation of methods for the determination of tortuosity of Li-ion battery separators

The porosities and tortuosities are commonly utilized to characterize the microstructure of a Li-ion battery's separator and are adopted as key input parameters in advanced battery models. Herein, a general classification of the tortuosity for a porous medium is introduced based on its bi-fold significance, i.e., the geometrical and physical tortuosities. Then, three different methods for the determination of separator's electrical tortuosity are introduced and compared, which include the empirical Bruggeman equation, the experimental method using Electrochemical Impedance Spectrum (EIS) testing a the numerical method using realistic 3D microstructure of the separator obtained from nanoscale X-ray Computed Tomography (XCT). In addition, the connection between the geometrical tortuosity and the electrical tortuosity of a separator is established by introducing the electrical phenomenological factor (\b{eta}_e), which can facilitate the understanding of the relationship between the microstructure characteristics and transport properties of the separators. Furthermore, to quantitively compare the values of the tortuosities determined by different methods, the corresponding effective transport coefficients (δ) are compared, which was usually used as a correction for effective diffusivity and conductivity of electrolytes in porous media.

physics.app-ph↗

arXiv2022

Dissecting Tensor Cores via Microbenchmarks: Latency, Throughput and Numeric Behaviors

Tensor Cores have been an important unit to accelerate Fused Matrix Multiplication Accumulation (MMA) in all NVIDIA GPUs since Volta Architecture. To program Tensor Cores, users have to use either legacy wmma APIs or current mma APIs. Legacy wmma APIs are more easy-to-use but can only exploit limited features and power of Tensor Cores. Specifically, wmma APIs support fewer operand shapes and can not leverage the new sparse matrix multiplication feature of the newest Ampere Tensor Cores. However, the performance of current programming interface has not been well explored. Furthermore, the computation numeric behaviors of low-precision floating points (TF32, BF16, and FP16) supported by the newest Ampere Tensor Cores are also mysterious. In this paper, we explore the throughput and latency of current programming APIs. We also intuitively study the numeric behaviors of Tensor Cores MMA and profile the intermediate operations including multiplication, addition of inner product, and accumulation. All codes used in this work can be found in https://github.com/sunlex0717/DissectingTensorCores.

cs.AR↗

arXiv2022

Representation Learning for Continuous Action Spaces is Beneficial for Efficient Policy Learning

Deep reinforcement learning (DRL) breaks through the bottlenecks of traditional reinforcement learning (RL) with the help of the perception capability of deep learning and has been widely applied in real-world problems.While model-free RL, as a class of efficient DRL methods, performs the learning of state representations simultaneously with policy learning in an end-to-end manner when facing large-scale continuous state and action spaces. However, training such a large policy model requires a large number of trajectory samples and training time. On the other hand, the learned policy often fails to generalize to large-scale action spaces, especially for the continuous action spaces. To address this issue, in this paper we propose an efficient policy learning method in latent state and action spaces. More specifically, we extend the idea of state representations to action representations for better policy generalization capability. Meanwhile, we divide the whole learning task into learning with the large-scale representation models in an unsupervised manner and learning with the small-scale policy model in the RL manner.The small policy model facilitates policy learning, while not sacrificing generalization and expressiveness via the large representation model. Finally,the effectiveness of the proposed method is demonstrated by MountainCar,CarRacing and Cheetah experiments.

cs.LG↗

arXiv2022

Users are Closer than they Appear: Protecting User Location from WiFi APs

WiFi-based indoor localization has now matured for over a decade. Most of the current localization algorithms rely on the WiFi access points (APs) in the enterprise network to localize the WiFi user accurately. Thus, the WiFi user's location information could be easily snooped by an attacker listening through a compromised WiFi AP. With indoor localization and navigation being the next step towards automation, it is important to give users the capability to defend against such attacks. In this paper, we present MIRAGE, a system that can utilize the downlink physical layer information to create a defense against an attacker snooping on a WiFi user's location information. MIRAGE achieves this by utilizing the beamforming capability of the transmitter that is already part of the WiFi protocols. With this initial idea, we have demonstrated that the user can obfuscate his/her location from the WiFi AP always with no compromise to the throughput of the existing WiFi communication system and reduce the user location accuracy of the attacker from 2.3m to more than 10m.

cs.CR↗

arXiv2022

DyOb-SLAM : Dynamic Object Tracking SLAM System

Simultaneous Localization & Mapping (SLAM) is the process of building a mutual relationship between localization and mapping of the subject in its surrounding environment. With the help of different sensors, various types of SLAM systems have developed to deal with the problem of building the relationship between localization and mapping. A limitation in the SLAM process is the lack of consideration of dynamic objects in the mapping of the environment. We propose the Dynamic Object Tracking SLAM (DyOb-SLAM), which is a Visual SLAM system that can localize and map the surrounding dynamic objects in the environment as well as track the dynamic objects in each frame. With the help of a neural network and a dense optical flow algorithm, dynamic objects and static objects in an environment can be differentiated. DyOb-SLAM creates two separate maps for both static and dynamic contents. For the static features, a sparse map is obtained. For the dynamic contents, a trajectory global map is created as output. As a result, a frame to frame real-time based dynamic object tracking system is obtained. With the pose calculation of the dynamic objects and camera, DyOb-SLAM can estimate the speed of the dynamic objects with time. The performance of DyOb-SLAM is observed by comparing it with a similar Visual SLAM system, VDO-SLAM and the performance is measured by calculating the camera and object pose errors as well as the object speed error.

cs.RO↗