arXiv subjects

Kun Hu

Publications and source records attributed to Kun Hu.

At least 91 records · Page 5Linked to original sources

arXiv2023

The First X-ray Polarization Observation of the Black Hole X-ray Binary 4U 1630-47 in the Steep Power Law State

The Imaging X-ray Polarimetry Explorer (IXPE) observed the black hole X-ray binary 4U 1630-47 in the steep power law (or very high) state. The observations reveal a linear polarization degree of the 2-8 keV X-rays of 6.8 +/- 0.2 % at a position angle of 21°.3 +/- 0°.9 East of North (all errors at 1σ confidence level). Whereas the polarization degree increases with energy, the polarization angle stays constant within the accuracy of our measurements. We compare the polarization of the source in the steep power-law state with the previous IXPE measurement of the source in the high soft state. We find that even though the source flux and spectral shape are significantly different between the high soft state and the steep power-law state, their polarization signatures are similar. Assuming that the polarization of both the thermal and power-law emission components are constant over time, we estimate the power-law component polarization to be 6.8-7.0% and note that the polarization angle of the thermal and power-law components must be approximately aligned. We discuss the implications for the origin of the power-law component and the properties of the emitting plasma.

astro-ph.HE↗

arXiv2023

Continuous Intermediate Token Learning with Implicit Motion Manifold for Keyframe Based Motion Interpolation

Deriving sophisticated 3D motions from sparse keyframes is a particularly challenging problem, due to continuity and exceptionally skeletal precision. The action features are often derivable accurately from the full series of keyframes, and thus, leveraging the global context with transformers has been a promising data-driven embedding approach. However, existing methods are often with inputs of interpolated intermediate frame for continuity using basic interpolation methods with keyframes, which result in a trivial local minimum during training. In this paper, we propose a novel framework to formulate latent motion manifolds with keyframe-based constraints, from which the continuous nature of intermediate token representations is considered. Particularly, our proposed framework consists of two stages for identifying a latent motion subspace, i.e., a keyframe encoding stage and an intermediate token generation stage, and a subsequent motion synthesis stage to extrapolate and compose motion data from manifolds. Through our extensive experiments conducted on both the LaFAN1 and CMU Mocap datasets, our proposed method demonstrates both superior interpolation accuracy and high visual similarity to ground truth motions.

cs.CV↗

arXiv2023

Multi-Scale Control Signal-Aware Transformer for Motion Synthesis without Phase

Synthesizing controllable motion for a character using deep learning has been a promising approach due to its potential to learn a compact model without laborious feature engineering. To produce dynamic motion from weak control signals such as desired paths, existing methods often require auxiliary information such as phases for alleviating motion ambiguity, which limits their generalisation capability. As past poses often contain useful auxiliary hints, in this paper, we propose a task-agnostic deep learning method, namely Multi-scale Control Signal-aware Transformer (MCS-T), with an attention based encoder-decoder architecture to discover the auxiliary information implicitly for synthesizing controllable motion without explicitly requiring auxiliary information such as phase. Specifically, an encoder is devised to adaptively formulate the motion patterns of a character's past poses with multi-scale skeletons, and a decoder driven by control signals to further synthesize and predict the character's state by paying context-specialised attention to the encoded past motion patterns. As a result, it helps alleviate the issues of low responsiveness and slow transition which often happen in conventional methods not using auxiliary information. Both qualitative and quantitative experimental results on an existing biped locomotion dataset, which involves diverse types of motion transitions, demonstrate the effectiveness of our method. In particular, MCS-T is able to successfully generate motions comparable to those generated by the methods using auxiliary information.

cs.CV↗

arXiv2023

Robust Knowledge Adaptation for Federated Unsupervised Person ReID

Person Re-identification (ReID) has been extensively studied in recent years due to the increasing demand in public security. However, collecting and dealing with sensitive personal data raises privacy concerns. Therefore, federated learning has been explored for Person ReID, which aims to share minimal sensitive data between different parties (clients). However, existing federated learning based person ReID methods generally rely on laborious and time-consuming data annotations and it is difficult to guarantee cross-domain consistency. Thus, in this work, a federated unsupervised cluster-contrastive (FedUCC) learning method is proposed for Person ReID. FedUCC introduces a three-stage modelling strategy following a coarse-to-fine manner. In detail, generic knowledge, specialized knowledge and patch knowledge are discovered using a deep neural network. This enables the sharing of mutual knowledge among clients while retaining local domain-specific knowledge based on the kinds of network layers and their parameters. Comprehensive experiments on 8 public benchmark datasets demonstrate the state-of-the-art performance of our proposed method.

cs.CV↗

arXiv2022

Artificial Intelligence Security Competition (AISC)

The security of artificial intelligence (AI) is an important research area towards safe, reliable, and trustworthy AI systems. To accelerate the research on AI security, the Artificial Intelligence Security Competition (AISC) was organized by the Zhongguancun Laboratory, China Industrial Control Systems Cyber Emergency Response Team, Institute for Artificial Intelligence, Tsinghua University, and RealAI as part of the Zhongguancun International Frontier Technology Innovation Competition (https://www.zgc-aisc.com/en). The competition consists of three tracks, including Deepfake Security Competition, Autonomous Driving Security Competition, and Face Recognition Security Competition. This report will introduce the competition rules of these three tracks and the solutions of top-ranking teams in each track.

cs.CR↗

arXiv2022

TLDW: Extreme Multimodal Summarisation of News Videos

Multimodal summarisation with multimodal output is drawing increasing attention due to the rapid growth of multimedia data. While several methods have been proposed to summarise visual-text contents, their multimodal outputs are not succinct enough at an extreme level to address the information overload issue. To the end of extreme multimodal summarisation, we introduce a new task, eXtreme Multimodal Summarisation with Multimodal Output (XMSMO) for the scenario of TL;DW - Too Long; Didn't Watch, akin to TL;DR. XMSMO aims to summarise a video-document pair into a summary with an extremely short length, which consists of one cover frame as the visual summary and one sentence as the textual summary. We propose a novel unsupervised Hierarchical Optimal Transport Network (HOT-Net) consisting of three components: hierarchical multimodal encoders, hierarchical multimodal fusion decoders, and optimal transport solvers. Our method is trained, without using reference summaries, by optimising the visual and textual coverage from the perspectives of the distance between the semantic distributions under optimal transport plans. To facilitate the study on this task, we collect a large-scale dataset XMSMO-News by harvesting 4,891 video-document pairs. The experimental results show that our method achieves promising performance in terms of ROUGE and IoU metrics.

cs.CV↗

arXiv2022

High-Energy Photon Opacity in the Twisted Magnetospheres of Magnetars

Magnetars are neutron stars characterized by strong surface magnetic fields generally exceeding the quantum critical value of 44.1 TeraGauss. High-energy photons propagating in their magnetospheres can be attenuated by QED processes like photon splitting and magnetic pair creation. In this paper, we compute the opacities due to photon splitting and pair creation by photons emitted anywhere in the magnetosphere of a magnetar. Axisymmetric, twisted dipole field configurations embedded in the Schwarzschild metric are treated. The paper computes the maximum energies for photon transparency that permit propagation to infinity in curved spacetime. Special emphasis is given to cases where photons are generated along magnetic field loops and/or in polar regions; these cases directly relate to resonant inverse Compton scattering models for the hard X-ray emission from magnetars and Comptonized soft gamma-ray emission from giant flares. We find that increases in magnetospheric twists raise or lower photon opacities, depending on both the emission locale, and the competition between field line straightening and field strength enhancement. Consequently, given the implicit spectral transparency of hard X-ray bursts and persistent "tail" emission of magnetars, photon splitting considerations constrain their emission region locales and the twist angle of the magnetosphere; these constraints can be probed by future soft gamma-ray telescopes such as COSI and AMEGO. The inclusion of twists generally increases the opaque volume of pair creation by photons above its threshold, except when photons are emitted in polar regions and approximately parallel to the field.

astro-ph.HE↗

arXiv2022

Multi-level Adversarial Spatio-temporal Learning for Footstep Pressure based FoG Detection

Freezing of gait (FoG) is one of the most common symptoms of Parkinson's disease, which is a neurodegenerative disorder of the central nervous system impacting millions of people around the world. To address the pressing need to improve the quality of treatment for FoG, devising a computer-aided detection and quantification tool for FoG has been increasingly important. As a non-invasive technique for collecting motion patterns, the footstep pressure sequences obtained from pressure sensitive gait mats provide a great opportunity for evaluating FoG in the clinic and potentially in the home environment. In this study, FoG detection is formulated as a sequential modelling task and a novel deep learning architecture, namely Adversarial Spatio-temporal Network (ASTN), is proposed to learn FoG patterns across multiple levels. A novel adversarial training scheme is introduced with a multi-level subject discriminator to obtain subject-independent FoG representations, which helps to reduce the over-fitting risk due to the high inter-subject variance. As a result, robust FoG detection can be achieved for unseen subjects. The proposed scheme also sheds light on improving subject-level clinical studies from other scenarios as it can be integrated with many existing deep architectures. To the best of our knowledge, this is one of the first studies of footstep pressure-based FoG detection and the approach of utilizing ASTN is the first deep neural network architecture in pursuit of subject-independent representations. Experimental results on 393 trials collected from 21 subjects demonstrate encouraging performance of the proposed ASTN for FoG detection with an AUC 0.85.

cs.CV↗

arXiv2022

ADM formulation and Hamiltonian analysis of $f(Q)$ gravity

$f(Q)$ gravity is an extension of the symmetric teleparallel equivalent to general relativity. We demonstrate the Hamiltonian analysis of $f(Q)$ gravity with fixing the coincident gauge condition. Using the standard Dirac-Bergmann algorithm, we show that $f(Q)$ gravity has 8 physical degrees of freedom. This result reflects that the diffeomorphism symmetry of $f(Q)$ gravity is completely broken due to the gauge fixing. Moreover, in terms of the perturbations, we discuss the possible mode decomposition of these degrees of freedom.

gr-qc↗

arXiv2022

M2-Net: Multi-stages Specular Highlight Detection and Removal in Multi-scenes

In this paper, we propose a novel uniformity framework for highlight detection and removal in multi-scenes, including synthetic images, face images, natural images, and text images. The framework consists of three main components, highlight feature extractor module, highlight coarse removal module, and highlight refine removal module. Firstly, the highlight feature extractor module can directly separate the highlight feature and non-highlight feature from the original highlight image. Then highlight removal image is obtained using a coarse highlight removal network. To further improve the highlight removal effect, the refined highlight removal image is finally obtained using refine highlight removal module based on contextual highlight attention mechanisms. Extensive experimental results in multiple scenes indicate that the proposed framework can obtain excellent visual effects of highlight removal and achieve state-of-the-art results in several quantitative evaluation metrics. Our algorithm is applied for the first time in video highlight removal with promising results.

cs.CV↗

arXiv2022

Polarized Radiation Signals from Highly Magnetized Neutron Star Surfaces

The surfaces of neutron stars are likely sources of strongly polarized soft X rays due to the presence of strong magnetic fields. Scattering transport in the surface layers is critical to the determination of the emergent anisotropy of light intensity, and is strongly influenced by the complicated interplay between linear and circular polarization information. We have developed a magnetic Thomson scattering simulation to model the outer layers of fully-ionized atmospheres in such compact objects. Here we summarize emergent intensities and polarizations from extended atmospheric simulations, spanning considerable ranges of magnetic colatitudes. General relativistic propagation of light from the surface to infinity is fully included. The net polarization degrees are moderate and not very small when summing over a variety of field directions. These results provide an important foundation for observations of magnetars to be acquired by NASA's new IXPE X-ray polarimeter and future X-ray polarimetry missions.

astro-ph.HE↗

arXiv2022

OTExtSum: Extractive Text Summarisation with Optimal Transport

Extractive text summarisation aims to select salient sentences from a document to form a short yet informative summary. While learning-based methods have achieved promising results, they have several limitations, such as dependence on expensive training and lack of interpretability. Therefore, in this paper, we propose a novel non-learning-based method by for the first time formulating text summarisation as an Optimal Transport (OT) problem, namely Optimal Transport Extractive Summariser (OTExtSum). Optimal sentence extraction is conceptualised as obtaining an optimal summary that minimises the transportation cost to a given document regarding their semantic distributions. Such a cost is defined by the Wasserstein distance and used to measure the summary's semantic coverage of the original document. Comprehensive experiments on four challenging and widely used datasets - MultiNews, PubMed, BillSum, and CNN/DM demonstrate that our proposed method outperforms the state-of-the-art non-learning-based methods and several recent learning-based methods in terms of the ROUGE metric.

cs.CL↗

arXiv2022

Thermodynamical and topological properties of metastable Fe3Sn

Combining experimental data, first-principles calculations, and Calphad assessment, thermodynamic and topological transport properties of the Fe-Sn system were investigated. Density functional theory (DFT) calculations were performed to evaluate the intermetallics' finite-temperature heat capacity (Cp). A consistent thermodynamic assessment of the Fe-Sn phase diagram was achieved by using the experimental and DFT results, together with all available data from previous publications. Hence, the metastable phase Fe3Sn was firstly introduced into the current metastable phase diagram, and corrected phase locations of Fe5Sn3 and Fe3Sn2 under the newly measured corrected temperature ranges. Furthermore, the anomalous Hall conductivity and anomalous Nernst conductivity of Fe3Sn were calculated, with magnetization directions and doping considered as perturbations to tune such transport properties. It was observed that the enhanced anomalous Hall and Nernst conductivities originate from the combination of nodal lines and small gap areas that can be tuned by doping Mn at Fe sites and varying magnetization direction

cond-mat.mtrl-sci↗

arXiv2022

Intensity and Polarization Characteristics of Extended Neutron Star Surface Regions

The surfaces of neutron stars are sources of strongly polarized soft X rays due to the presence of strong magnetic fields. Radiative transfer mediated by electron scattering and free-free absorption is central to defining local surface anisotropy and polarization signatures. Scattering transport is strongly influenced by the complicated interplay between linear and circular polarizations. This complexity has been captured in a sophisticated magnetic Thomson scattering simulation we recently developed to model the outer layers of fully-ionized atmospheres in such compact objects, heretofore focusing on case studies of localized surface regions. Yet, the interpretation of observed intensity pulse profiles and their efficacy in constraining key neutron star geometry parameters is critically dependent upon adding up emission from extended surface regions. In this paper, intensity, anisotropy and polarization characteristics from such extended atmospheres, spanning considerable ranges of magnetic colatitudes, are determined using our transport simulation. These constitute a convolution of varied properties of Stokes parameter information at disparate surface locales with different magnetic field strengths and directions relative to the local zenith. Our analysis includes full general relativistic propagation of light from the surface to an observer at infinity. The array of pulse profiles for intensity and polarization presented highlights how powerful probes of stellar geometry are possible. Significant phase-resolved polarization degrees in the range of 10-60% are realized when summing over a variety of surface field directions. These results provide an important background for observations to be acquired by NASA's new IXPE X-ray polarimetry mission.

astro-ph.HE↗

arXiv2021

Sign Language Translation with Hierarchical Spatio-TemporalGraph Neural Network

Sign language translation (SLT), which generates text in a spoken language from visual content in a sign language, is important to assist the hard-of-hearing community for their communications. Inspired by neural machine translation (NMT), most existing SLT studies adopted a general sequence to sequence learning strategy. However, SLT is significantly different from general NMT tasks since sign languages convey messages through multiple visual-manual aspects. Therefore, in this paper, these unique characteristics of sign languages are formulated as hierarchical spatio-temporal graph representations, including high-level and fine-level graphs of which a vertex characterizes a specified body part and an edge represents their interactions. Particularly, high-level graphs represent the patterns in the regions such as hands and face, and fine-level graphs consider the joints of hands and landmarks of facial regions. To learn these graph patterns, a novel deep learning architecture, namely hierarchical spatio-temporal graph neural network (HST-GNN), is proposed. Graph convolutions and graph self-attentions with neighborhood context are proposed to characterize both the local and the global graph properties. Experimental results on benchmark datasets demonstrated the effectiveness of the proposed method.

cs.CV↗

arXiv2020

The Mysterious Magnetospheres of Magnetars

Magnetars are the most luminous compact objects in the stellar mass range observed in the Milky Way, with giant flares of hard X-ray power ~10^45 erg/sec being detected from three soft gamma repeaters in the Galactic neighborhood. Periodicity seen in magnetar persistent emission, and a distinctive "spin-down" lengthening of this period, have driven the paradigm that strongly-magnetized neutron stars constitute these fascinating sources. The steady X-ray emission includes both thermal atmospheric components, and magnetospheric contributions that are manifested as hard X-ray "tails." This paper addresses observational and theoretical elements pertinent to the steady hard X-ray emission of magnetars, focusing on dissipative processes in their magnetospheres, and elements of Comptonization and polarization. It also discusses the action and possible signatures of the exotic and fundamental QED mechanisms of photon splitting and magnetic pair creation, and the quest for their observational vindication.

astro-ph.HE↗

arXiv2020

Polarized Radiation Transfer in Neutron Star Surface Layers

The study of polarized radiation transfer in the highly-magnetized surface locales of neutron stars is of great interest to the understanding of accreting X-ray pulsars, rotation-powered pulsars and magnetars. This paper explores scattering transport in the classical magnetic Thomson domain that is of broad applicability to these neutron star classes. The development of a Monte Carlo simulation for the polarized radiative transfer is detailed: it employs an electric field vector formalism to enable a breadth of utility in relating linear, circular and elliptical polarizations. The simulation can be applied to any neutron star surface locale, and is adaptable to accretion column and magnetospheric problems. Validation of the code for both intensity and Stokes parameter determination is illustrated in a variety of ways. Representative results for emergent polarization signals from surface layers are presented for both polar and equatorial magnetic locales, exhibiting contrasting signatures between the two regions. There is also a strong dependence of these characteristics on the ratio of the frequency of a photon to the cyclotron frequency. Polarization signatures for high opacity domains are presented, highlighting compact analytic approximations for the Stokes parameters and anisotropy relative to the local field direction for an extended range of frequencies. These are very useful in defining injection conditions deep in the simulation slab geometries, expediting the generation of emission signals from highly opaque stellar atmospheres. The results are interpreted throughout using the polarization characteristics of the magnetic Thomson differential cross section.

astro-ph.HE↗

arXiv2019

Opacities for Photon Splitting and Pair Creation in Neutron Star Magnetospheres

Over the last four decades, persistent and flaring emission of magnetars observed by various telescopes has provided us with a suite of light curves and spectra in soft and hard X-rays, with no emission yet detected above around 1 MeV. Attenuation of such high-energy photons by magnetic pair creation and photon splitting is expected to be active in the magnetospheres of magnetars, possibly accounting for the paucity of gamma-rays in their signals. This paper explores polarization-dependent opacities for these two QED processes in static vacuum dipole magnetospheres of highly-magnetized neutron stars, calculating attenuation lengths and determining escape energies, which are the maximum photon energies for transparency out to infinity. The numerical trajectory integral analysis in flat and curved spacetimes provides upper bounds of a few MeV or less to the visible energies for magnetars for locales proximate to the stellar surface. Photon splitting opacity alone puts constraints on the possible emission locales in their magnetospheres: regions within field loops of maximum altitudes 2-4 stellar radii are not commensurate with maximum detected energies of around 250 keV. These constraints apply not only to magnetar flares but also to their quiescent hard X-ray tail emission. An exploration of photon splitting attenuation in the context of a resonant inverse Compton scattering model for the hard X-ray tails derives distinctive phase-resolved spectroscopic and polarimetric signatures, of significant interest for future MeV-band missions such as AMEGO and e-ASTROGAM.

astro-ph.HE↗