arXiv subjects

Lei Hu

Publications and source records attributed to Lei Hu.

At least 91 records · Page 5Linked to original sources

arXiv2023

A large-scale dataset for end-to-end table recognition in the wild

Table recognition (TR) is one of the research hotspots in pattern recognition, which aims to extract information from tables in an image. Common table recognition tasks include table detection (TD), table structure recognition (TSR) and table content recognition (TCR). TD is to locate tables in the image, TCR recognizes text content, and TSR recognizes spatial ogical structure. Currently, the end-to-end TR in real scenarios, accomplishing the three sub-tasks simultaneously, is yet an unexplored research area. One major factor that inhibits researchers is the lack of a benchmark dataset. To this end, we propose a new large-scale dataset named Table Recognition Set (TabRecSet) with diverse table forms sourcing from multiple scenarios in the wild, providing complete annotation dedicated to end-to-end TR research. It is the largest and first bi-lingual dataset for end-to-end TR, with 38.1K tables in which 20.4K are in English\, and 17.7K are in Chinese. The samples have diverse forms, such as the border-complete and -incomplete table, regular and irregular table (rotated, distorted, etc.). The scenarios are multiple in the wild, varying from scanned to camera-taken images, documents to Excel tables, educational test papers to financial invoices. The annotations are complete, consisting of the table body spatial annotation, cell spatial logical annotation and text content for TD, TSR and TCR, respectively. The spatial annotation utilizes the polygon instead of the bounding box or quadrilateral adopted by most datasets. The polygon spatial annotation is more suitable for irregular tables that are common in wild scenarios. Additionally, we propose a visualized and interactive annotation tool named TableMe to improve the efficiency and quality of table annotation.

cs.CV↗

arXiv2023

Reconfigurable Intelligent Surface-aided Secret Key Generation in Multi-Cell Systems

Physical-layer key generation (PKG) exploits the reciprocity and randomness of wireless channels to generate a symmetric key between two legitimate communication ends. However, in multi-cell systems, PKG suffers from severe pilot contamination due to the reuse of pilots in different cells. In this paper, we invoke multiple reconfigurable intelligent surfaces (RISs) for adaptively shaping the environment and enhancing the PKG performance. To this end, we formulate an optimization problem to maximize the weighted sum key rate (WSKR) by jointly optimizing the precoding matrices at the base stations (BSs) and the phase shifts at the RISs. For addressing the non-convexity of the problem, we derive an upper bound of the WSKR and prove its tightness. To tackle the upper bound maximization problem, we apply an alternating optimization (AO)-based algorithm to divide the joint optimization into two sub-problems. We apply the Lagrangian dual approach based on the Karush-Kuhn-Tucker (KKT) conditions for the sub-problem of precoding matrices and adopt a projected gradient ascent (PGA) algorithm for the sub-problem of phase shifts. Simulation results confirm the near-optimal performance of the proposed algorithm and the effectiveness of RISs for improving the WSKR via mitigating pilot contamination.

cs.IT↗

arXiv2023

Kirigami-Inspired Thermal Regulator

One of the current challenges in nanoscience is tailoring phononic devices, such as thermal regulators and thermal computing. This has long been a rather elusive task because the thermal-switching ratio is not as high as electronic analogs. Mapping from a topological kirigami assembly, nitrogen-doped porous graphene metamaterials on the nanoscale are inversely designed with a thermal-switching ratio of 27.79, which is more than double the value of previous work. We trace this behavior to the chiral folding-unfolding deformation, resulting in a metal-insulator transition. This study provides a nanomaterial design paradigm to bridge the gap between kinematics and functional metamaterials that motivates the development of high-performance thermal regulators.

cond-mat.mtrl-sci↗

arXiv2022

Prospects of Searching for Type Ia Supernovae with 2.5-m Wide Field Survey Telescope

Type Ia Supernovae (SNe Ia) are the thermonuclear explosion of a carbon-oxygen white dwarf (WD) and are well-known as a distance indicator. However, it is still unclear how WDs increase their mass near the Chandrasekhar limit and how the thermonuclear runaway happens. The observational clues associated with these open questions, such as the photometric data within hours to days since the explosion, are scarce. Thus, an essential way is to discover SNe Ia at specific epochs with optimal surveys. The 2.5-m Wide Field Survey Telescope (WFST) is an upcoming survey facility deployed in western China. In this paper, we assess the detecability of SNe Ia with mock observations of WFST. Followed by the volumetric rate, we generate a spectral series of SNe Ia based on a data-based model and introduce the line-of-sight extinction to calculate the brightness from the observer. By comparing with the detection limit of WFST, which is affected by the observing conditions, we can count the number of SNe Ia discovered by mock WFST observations. We expect that WFST can find more than $3.0\times10^{4}$ pre-maximum SNe Ia within one-year running. In particular, WFST could discover about 45 bright SNe Ia, 99 early-phase SNe Ia, or $1.1\times10^{4}$ well-observed SNe Ia with the hypothesized Wide, Deep, or Medium mode, respectively, suggesting WFST will be an influential facility in time-domain astronomy.

astro-ph.SR↗

arXiv2022

Deep Drilling in the Time Domain with DECam: Survey Characterization

This paper presents a new optical imaging survey of four deep drilling fields (DDFs), two Galactic and two extragalactic, with the Dark Energy Camera (DECam) on the 4 meter Blanco telescope at the Cerro Tololo Inter-American Observatory (CTIO). During the first year of observations in 2021, $>$4000 images covering 21 square degrees (7 DECam pointings), with $\sim$40 epochs (nights) per field and 5 to 6 images per night per filter in $g$, $r$, $i$, and/or $z$, have become publicly available (the proprietary period for this program is waived). We describe the real-time difference-image pipeline and how alerts are distributed to brokers via the same distribution system as the Zwicky Transient Facility (ZTF). In this paper, we focus on the two extragalactic deep fields (COSMOS and ELAIS-S1), characterizing the detected sources and demonstrating that the survey design is effective for probing the discovery space of faint and fast variable and transient sources. We describe and make publicly available 4413 calibrated light curves based on difference-image detection photometry of transients and variables in the extragalactic fields. We also present preliminary scientific analysis regarding Solar System small bodies, stellar flares and variables, Galactic anomaly detection, fast-rising transients and variables, supernovae, and active galactic nuclei.

astro-ph.IM↗

arXiv2022

Artificial Intelligence Assisted Inversion (AIAI): Quantifying the Spectral Features of $^{56}$Ni of Type Ia Supernovae

Following our previous study of Artificial Intelligence Assisted Inversion (AIAI) of supernova analyses (Chen et al. 2020), we train a set of deep neural networks based on the one-dimensional radiative transfer code TARDIS (Kerzendorf & Sim 2014) to simulate the optical spectra of Type Ia supernovae (SNe Ia) between 10 and 40 days after the explosion. The neural networks are applied to derive the mass of 56Ni in velocity ranges well above the photosphere for a sample of 153 well-observed SNe Ia. Many SNe have multi-epoch observations for which the decay of the radioactive 56Ni can be tested quantitatively. The 56Ni mass derived from AIAI using the observed spectra as input for the sample is found to agree with the theoretical 56Ni decay rate. The AIAI reveals a spectral signature near 3890 Åwhich can be identified as being produced by multiple Ni II lines between 3950 and 4100 Å. The mass deduced from AIAI is correlated to the light-curve shapes of SNe Ia, with the SNe Ia with broader light curves showing larger 56Ni mass in the envelope. AIAI enables spectral data of SNe to be quantitatively analyzed under theoretical frameworks based on well-defined physical assumptions.

astro-ph.HE↗

arXiv2022

The Interaction of Supernova 2018evt with a Substantial Amount of Circumstellar Matter -- An SN1997cy-like Event

A rare class of supernovae (SNe) is characterized by strong interaction between the ejecta and several solar masses of circumstellar matter (CSM) as evidenced by strong Balmer-line emission. Within the first few weeks after the explosion, they may display spectral features similar to overluminous Type Ia SNe, while at later phase their observation properties exhibit remarkable similarities with some extreme case of Type IIn SNe that show strong Balmer lines years after the explosion. We present polarimetric observations of SN2018evt obtained by the ESO Very Large Telescope from 172 to 219 days after the estimated time of peak luminosity to study the geometry of the CSM. The nonzero continuum polarization decreases over time, suggesting that the mass loss of the progenitor star is aspherical. The prominent H$α$ emission can be decomposed into a broad, time-evolving component and an intermediate-width, static component. The former shows polarized signals, and it is likely to arise from a cold dense shell (CDS) within the region between the forward and reverse shocks. The latter is significantly unpolarized, and it is likely to arise from shocked, fragmented gas clouds in the H-rich CSM. We infer that SN2018evt exploded inside a massive and aspherical circumstellar cloud. The symmetry axes of the CSM and the SN appear to be similar. SN\,2018evt shows observational properties common to events that display strong interaction between the ejecta and CSM, implying that they share similar circumstellar configurations. Our preliminary estimate also suggests that the circumstellar environment of SN2018evt has been significantly enriched at a rate of $\sim0.1$ M$_\odot$ yr$^{-1}$ over a period of $>100$ yr.

astro-ph.HE↗

arXiv2022

Reconfigurable Intelligent Surface-Assisted Secret Key Generation in Spatially Correlated Channels

Reconfigurable intelligent surface (RIS) is a disruptive technology to enhance the performance of physical-layer key generation (PKG) thanks to its ability to smartly customize the radio environments. Existing RIS-assisted PKG methods are mainly based on the idealistic assumption of an independent and identically distributed (i.i.d.) channel model at both the base station (BS) and the RIS. However, the i.i.d. model is inaccurate for a typical RIS in an isotropic scattering environment and neglecting the existence of channel spatial correlation would possibly degrade the PKG performance. In this paper, we establish a general spatially correlated channel model and propose a new channel probing framework based on the transmit and the reflective beamforming. We derive a closed-form key generation rate (KGR) expression and formulate an optimization problem, which is solved by using the low-complexity Block Successive Upper-bound Minimization (BSUM) with Mirror-Prox method. Simulation results show that compared to the existing methods based on the i.i.d. fading model, our proposed method achieves about $5$ dB transmit power gain when the spacing between two neighboring RIS elements is a quarter of the wavelength. Also, the KGR increases significantly with the number of RIS elements while that increases marginally with the number of BS antennas.

cs.IT↗

arXiv2022

Neural3Points: Learning to Generate Physically Realistic Full-body Motion for Virtual Reality Users

Animating an avatar that reflects a user's action in the VR world enables natural interactions with the virtual environment. It has the potential to allow remote users to communicate and collaborate in a way as if they met in person. However, a typical VR system provides only a very sparse set of up to three positional sensors, including a head-mounted display (HMD) and optionally two hand-held controllers, making the estimation of the user's full-body movement a difficult problem. In this work, we present a data-driven physics-based method for predicting the realistic full-body movement of the user according to the transformations of these VR trackers and simulating an avatar character to mimic such user actions in the virtual world in real-time. We train our system using reinforcement learning with carefully designed pretraining processes to ensure the success of the training and the quality of the simulation. We demonstrate the effectiveness of the method with an extensive set of examples.

cs.GR↗

arXiv2022

Using 1991T/1999aa-like Type Ia Supernovae as Standardizable Candles

We present the photometry of 16 91T/99aa-like Type Ia Supernovae (SNe Ia) observed by the Las Cumbres Observatory. We also use an additional set of 21 91T/99aa-like SNe Ia and 87 normal SNe Ia from the literature for an analysis of the standardizability of the luminosity of 91T/99aa-like SNe. We find that 91T/99aa-like SNe are 0.2 mag brighter than normal SNe Ia, even when fully corrected by the light curve shapes and colors. The weighted root-mean-square of 91T/99aa-like SNe (with $z_{CMB}>0.01$) Hubble residuals is $0.25\pm0.03$ mag, suggesting that 91T/99aa-like SNe are also excellent relative distance indicators to $\pm$12%. We compare the Hubble residuals with the pseudo-equivalent width (pEW) of Si II $λλ$6355 around the date of maximum brightness. We find that there is a broken linear correlation in between those two measurements for our sample including both 91T/99aa-like and normal SNe Ia. As the $pEW_{max}$(Si II $λλ$6355) increasing, the Hubble residual increases when $pEW_{max}$(Si II $λλ$6355)$<55.6$ Å. However, the Hubble residual stays constant beyond this. Given that 91T/99aa-like SNe possess shallower Si II lines than normal SNe Ia, the linear correlation at $pEW_{max}$(Si II $λλ$6355)$<55.6$ Å can account for the overall discrepancy of Hubble residuals derived from the two subgroups. Such a systematic effect needs to be taken into account when using SNe Ia to measure luminosity distances.

astro-ph.HE↗

arXiv2022

Gravitationally Lensed Orphan Afterglows of Gamma-Ray Bursts

The cosmological nature of gamma-ray bursts (GRBs) implies that a small portion of them could be gravitationally lensed by foreground objects during their propagation. The gravitational lensing effect on the GRB prompt emission and on-axis afterglows has been discussed, and some candidates have been found in the literature. In this work, considering the high detection rate of GRB orphan afterglows in future wide-field survey era, we investigate the gravitationally lensed orphan afterglows in view of three lens models, i.e., the point-mass model, the singular isothermal sphere model, and the Chang-Refsdal model. The structure of the GRB jet itself is also incorporated in calculating the lensed afterglow light curves. It is found that lensed optical/X-ray orphan afterglows in principle could be diagnosed through their temporal characteristics, and the optical band is the best band to observe the galaxy-lensed orphan afterglows. Moreover, the event rate for galaxy-lensed orphan afterglows is estimated to be $\lesssim$ 1.8 $\text{yr}^{-1}$ for the whole sky. If most orphan afterglows could be identified (from other transients in the survey data), the optimistic detection rates of the 2.5m Wide Field Survey Telescope of China and 8.4m Vera Rubin Observatory Legacy Survey of Space and Time for galaxy-lensed orphan afterglows in the optical band are $\lesssim$ 0.01$\sim$0.02 $\text{yr}^{-1}$ and $\lesssim$ 0.04$\sim$0.08 $\text{yr}^{-1}$, respectively.

astro-ph.HE↗

arXiv2022

Joint Transmit and Reflective Beamforming for RIS-assisted Secret Key Generation

Reconfigurable intelligent surface (RIS) is a promising technique to enhance the performance of physical-layer key generation (PKG) due to its ability to smartly customize the radio environments. Existing RIS-assisted PKG methods are mainly based on the idealistic assumption of an independent and identically distributed (i.i.d.) channel model at both the transmitter and the RIS. However, the i.i.d. model is inaccurate for a typical RIS in an isotropic scattering environment. Also, neglecting the existence of channel spatial correlation would degrade the PKG performance. In this paper, we establish a general spatially correlated channel model in multi-antenna systems and propose a new PKG framework based on the transmit and the reflective beamforming at the base station (BS) and the RIS. Specifically, we derive a closed-form expression for characterizing the key generation rate (KGR) and obtain a globally optimal solution of the beamformers to maximize the KGR. Furthermore, we analyze the KGR performance difference between the one adopting the assumption of the i.i.d. model and that of the spatially correlated model. It is found that the beamforming designed for the correlated model outperforms that for the i.i.d. model while the KGR gain increases with the channel correlation. Simulation results show that compared to existing methods based on the i.i.d. fading model, our proposed method achieves about $5$ dB performance gain when the BS antenna correlation $ρ$ is $0.3$ and the RIS element spacing is half of the wavelength.

cs.IT↗

arXiv2022

Spatio-Temporal Gating-Adjacency GCN for Human Motion Prediction

Predicting future motion based on historical motion sequence is a fundamental problem in computer vision, and it has wide applications in autonomous driving and robotics. Some recent works have shown that Graph Convolutional Networks(GCN) are instrumental in modeling the relationship between different joints. However, considering the variants and diverse action types in human motion data, the cross-dependency of the spatio-temporal relationships will be difficult to depict due to the decoupled modeling strategy, which may also exacerbate the problem of insufficient generalization. Therefore, we propose the Spatio-Temporal Gating-Adjacency GCN(GAGCN) to learn the complex spatio-temporal dependencies over diverse action types. Specifically, we adopt gating networks to enhance the generalization of GCN via the trainable adaptive adjacency matrix obtained by blending the candidate spatio-temporal adjacency matrices. Moreover, GAGCN addresses the cross-dependency of space and time by balancing the weights of spatio-temporal modeling and fusing the decoupled spatio-temporal features. Extensive experiments on Human 3.6M, AMASS, and 3DPW demonstrate that GAGCN achieves state-of-the-art performance in both short-term and long-term predictions.

cs.CV↗

arXiv2022

Antarctic Survey Telescope 3-3: Overview, System Performance and Preliminary Observations at Yaoan, Yunnan

The third Antarctic Survey Telescope array instrument at Dome A in Antarctica, the AST3-3 telescope, has been in commissioning from March 2021. We deployed AST3-3 at the Yaoan astronomical station in Yunnan Province for an automatic time-domain survey and follow-up observations with an optimised observation and protection system. The telescope system of AST3-3 is similar to that of AST3-1 and AST3-2, except that it is equipped with a 14K~$ \times$~10K QHY411 CMOS camera. AST3-3 has a field of view of $1.65^\circ \times 1.23^\circ$ and is currently using the $g$ band filter. During commissioning at Yaoan, AST3-3 aims to conduct an extragalactic transient survey, coupled with prompt follow-ups of opportunity targets. In this paper, we present the architecture of the AST3-3 automatic observation system. We demonstrate the data processing of observations by representatives SN 2022eyw and GRB 210420B.

astro-ph.IM↗

arXiv2022

Image Subtraction in Fourier Space

Image subtraction is essential for transient detection in time-domain astronomy. The point spread function (PSF), photometric scaling, and sky background generally vary with time and across the field-of-view for imaging data taken with ground-based optical telescopes. Image subtraction algorithms need to match these variations for the detection of flux variability. An algorithm that can be fully parallelized is highly desirable for future time-domain surveys. Here we show the Saccadic Fast Fourier Transform (SFFT) algorithm for image differencing. SFFT uses $δ$-function basis for kernel decomposition, and the image subtraction is performed in Fourier Space. This brings about a remarkable improvement of computational performance of about an order of magnitude compared to other published image subtraction codes. SFFT can accommodate the spatial variations in wide-field imaging data, including PSF, photometric scaling, and sky background. However, the flexibility of the $δ$-function basis may also make it more prone to overfitting. The algorithm has been tested extensively in real astronomical data taken by a variety of telescopes. Moreover, the SFFT code allows for the spatial variations of the PSF and sky background to be fitted by spline functions.

astro-ph.IM↗

arXiv2022

Pipeline for Antarctic Survey Telescope 3-3 in Yaoan, Yunnan

AST3-3 is the third robotic facility of the Antarctic Survey Telescopes (AST3) for transient surveys to be deployed at Dome A, Antarctica. Due to the current pandemic, the telescope has been currently deployed at the Yaoan Observation Station in China, starting the commissioning observation and a transient survey. This paper presents a fully automatic data processing system for AST3-3 observations. The transient detection pipeline uses state-of-the-art image subtraction techniques optimised for GPU devices. Image reduction and transient photometry are accelerated by concurrent task methods. Our Python-based system allows for transient detection from wide-field data in a real-time and accurate way. A ResNet-based rotational-invariant neural network was employed to classify the transient candidates. As a result, the system enables auto-generation of transients and their light curves.

astro-ph.IM↗

arXiv2022

Spectroscopic Studies of Type Ia Supernovae Using LSTM Neural Networks

We present a data-driven method based on long short-term memory (LSTM) neural networks to analyze spectral time series of Type Ia supernovae (SNe Ia). The dataset includes 3091 spectra from 361 individual SNe Ia. The method allows for accurate reconstruction of the spectral sequence of an SN Ia based on a single observed spectrum around maximum light. The precision of the spectral reconstruction increases with more spectral time coverages, but the significant benefit of multiple epoch data at around optical maximum is only evident for observations separated by more than a week. The method shows great power in extracting the spectral information of SNe Ia, and suggests that the most critical information of an SN Ia can be derived from a single spectrum around the optical maximum. The algorithm we have developed is important for the planning of spectroscopic follow-up observations of future SN surveys with the LSST/Rubin and the WFIRST/Roman telescopes.

astro-ph.HE↗

arXiv2022

Reconfigurable Intelligent Surface for Physical Layer Key Generation: Constructive or Destructive?

Physical layer key generation (PKG) is a promising means to provide on-the-fly shared secret keys by exploiting the intrinsic randomness of the radio channel. However, the performance of PKG is highly dependent on the propagation environments. Due to its feature of controlling the wireless environment, reconfigurable intelligent surface~(RIS) is appealing to be applied in PKG. In this paper, in contrast to the existing literature, we investigate both the constructive and destructive effects of RIS on the PKG scheme. For the constructive aspect, we have identified static and wave-blockage environments as two RIS-empowered-PKG applications in future wireless systems. In particular, our experimental results in a static environment showed that RIS can enhance the entropy of the secret key, achieving a key generation rate (KGR) of 97.39 bit/s with a bit disagreement rate (BDR) of 0.083. In multi-user systems where some remote users are in worse channel conditions, the proposed RIS-assisted PKG algorithm improves the sum secret key rate by more than 2 dB, compared to the literature. Furthermore, we point out that RIS could be utilized by an attacker to perform new jamming and leakage attacks and give countermeasures, respectively. Finally, we outline future research directions for PKG systems in light of the RIS.

cs.IT↗