BZPEERReading, trust and collaboration for research
Menu

Discover

GraphFollow connections around a paper, topic or saved view.

Workspaces

HomePick up where your research flow left off.InboxHandle requests, replies and notifications from one place.CollectionsCurate reading lists, evidence sets and shareable dossiers.ResearchSee your private provenance graph, trail and imports.CollaborationMove from discovery into focused discussion rooms.PublishDraft, preview and publish full-text research articles.

Network

ExpertsFind specialists worth following or asking for help.CommunitiesJoin research circles organized around shared work.PartnersSee external organizations active around the same topics.

Opportunities

ListingsBrowse roles, calls and collaborations with clearer fit signals.

Account

Log inReturn to your reading and collaboration workspace.Create accountStart saving work, following people and joining teams.
Log inCreate account

Researcher profile

Le Zou

Le Zou contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
astro-ph.HEComputer Vision

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
8works
0followers
2topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

8 published item(s)

preprint2024arXiv

DAFD: Domain Adaptation via Feature Disentanglement for Image Classification

A good feature representation is the key to image classification. In practice, image classifiers may be applied in scenarios different from what they have been trained on. This so-called domain shift leads to a significant performance drop in image classification. Unsupervised domain adaptation (UDA) reduces the domain shift by transferring the knowledge learned from a labeled source domain to an unlabeled target domain. We perform feature disentanglement for UDA by distilling category-relevant features and excluding category-irrelevant features from the global feature maps. This disentanglement prevents the network from overfitting to category-irrelevant information and makes it focus on information useful for classification. This reduces the difficulty of domain alignment and improves the classification accuracy on the target domain. We propose a coarse-to-fine domain adaptation method called Domain Adaptation via Feature Disentanglement~(DAFD), which has two components: (1)the Category-Relevant Feature Selection (CRFS) module, which disentangles the category-relevant features from the category-irrelevant features, and (2)the Dynamic Local Maximum Mean Discrepancy (DLMMD) module, which achieves fine-grained alignment by reducing the discrepancy within the category-relevant features from different domains. Combined with the CRFS, the DLMMD module can align the category-relevant features properly. We conduct comprehensive experiment on four standard datasets. Our results clearly demonstrate the robustness and effectiveness of our approach in domain adaptive image classification tasks and its competitiveness to the state of the art.

0 citations0 reviewsNo code linkedOpen access
preprint2024arXiv

Domain-Invariant Proposals based on a Balanced Domain Classifier for Object Detection

Object recognition from images means to automatically find object(s) of interest and to return their category and location information. Benefiting from research on deep learning, like convolutional neural networks~(CNNs) and generative adversarial networks, the performance in this field has been improved significantly, especially when training and test data are drawn from similar distributions. However, mismatching distributions, i.e., domain shifts, lead to a significant performance drop. In this paper, we build domain-invariant detectors by learning domain classifiers via adversarial training. Based on the previous works that align image and instance level features, we mitigate the domain shift further by introducing a domain adaptation component at the region level within Faster \mbox{R-CNN}. We embed a domain classification network in the region proposal network~(RPN) using adversarial learning. The RPN can now generate accurate region proposals in different domains by effectively aligning the features between them. To mitigate the unstable convergence during the adversarial learning, we introduce a balanced domain classifier as well as a network learning rate adjustment strategy. We conduct comprehensive experiments using four standard datasets. The results demonstrate the effectiveness and robustness of our object detection approach in domain shift scenarios.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Early Evolution of a Newborn Magnetar with Strong Precession Motion in GRB 180620A

The observed early X-ray plateau in the afterglow lightcurves of some gamma-ray bursts (GRBs) is attributed to the dipole radiations (DRs) of a newborn magnetar. A quasi-periodic oscillation (QPO) signal in the plateau would be strong evidence of the magnetar precession motion. By making a time-frequency domain analysis for the X-ray afterglow lightcurve of GRB 180620A, we find a QPO signal of $\sim650$ seconds in its early X-ray plateau. We fit the lightcurve with a magnetar precession model by adopting the Markov chain Monte Carlo algorithm. The observed lightcurve and the QPO signal are well represented with our model. The derived magnetic field strength of the magnetar is $B_{\rm p}= (1.02^{+0.59}_{-0.61})\times10^{15}$~G. It rapidly spins down with angular velocity evolving as $Ω_{s} \propto(1+t/τ_{\rm sd})^{-0.96}$, where $τ_{\rm sd}=9430$~s. Its precession velocity evolution is even faster than $Ω_s$, i.e. $Ω_{ p}\propto (1+t/τ_{p})^{-2.18\pm0.11}$, where $τ_{p}=2239\pm206$~s. The inferred braking index is $n=2.04$. We argue that the extra energy loss via the magnetospheric processes results in its rapid spin-down, a low braking index of the magnetar, and the strong precession motion.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

New Benchmark for Household Garbage Image Recognition

Household garbage images are usually faced with complex backgrounds, variable illuminations, diverse angles, and changeable shapes, which bring a great difficulty in garbage image classification. Due to the ability to discover problem-specific features, deep learning and especially convolutional neural networks (CNNs) have been successfully and widely used for image representation learning. However, available and stable household garbage datasets are insufficient, which seriously limits the development of research and application. Besides, the state of the art in the field of garbage image classification is not entirely clear. To solve this problem, in this study, we built a new open benchmark dataset for household garbage image classification by simulating different lightings, backgrounds, angles, and shapes. This dataset is named 30 Classes of Household Garbage Images (HGI-30), which contains 18,000 images of 30 household garbage classes. The publicly available HGI-30 dataset allows researchers to develop accurate and robust methods for household garbage recognition. We also conducted experiments and performance analysis of the state-of-the-art deep CNN methods on HGI-30, which serves as baseline results on this benchmark.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

X-ray Flares Raising upon Magnetar Plateau as an Implication of a Surrounding Disk of Newborn Magnetized Neutron Star

The X-ray flares have usually been ascribed to long-lasting activities of the central engine of gamma-ray bursts (GRBs), e.g., fallback accretion. The GRB X-ray plateaus, however, favor a millisecond magnetar central engine. The fallback accretion can be significantly suppressed due to the propeller effect of a magnetar. Therefore, if the propeller regime cannot resist the mass flow onto the surface of the magnetar efficiently, the X-ray flares raise upon the magnetar plateau would be hinted. In this work, such peculiar cases are connected to the accretion process of a magnetar, and an implication for magnetar-disc structure is given. We investigate the repeating accretion process with multi-flare GRB 050730, and give a discussion for the accreting induced variation of the magnetic field in GRB 111209A. Two or more flares exhibit in the GRB 050730, GRB 060607A, and GRB 140304A; by adopting magnetar mass $M=1.4~ M_\odot$ and radius $R=12~\rm km$, the average mass flow rates of the corresponding surrounding disk are $3.53\times 10^{-4}~M_\odot~\rm s^{-1}$, $4.23\times 10^{-4}~M_\odot~\rm s^{-1}$, and $4.33\times 10^{-4}~M_\odot~\rm s^{-1}$, and the corresponding average sizes of the magnetosphere are $5.01~\rm \times10^{6} cm$, $6.45~\rm \times10^{6} cm$, and $1.09~\rm \times10^{7} cm$, respectively. A statistic analysis that contains 8 GRBs within 12 flares shows that the total mass loading in single flare is $\sim 2\times 10^{-5}~M_{\odot}$. In the lost mass of a disk, there are about 0.1% used to feed a collimated jet.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Early Optical Observations of GRB 150910A: Bright Jet Optical Afterglow and X-ray Dipole Radiation from a Magnetar Central Engine

Gamma-ray burst (GRB) 150910A was detected by {\it Swift}/BAT, and then rapidly observed by {\it Swift}/XRT, {\it Swift}/UVOT, and ground-based telescopes. We report Lick Observatory spectroscopic and photometric observations of GRB~150910A, and we investigate the physical origins of both the optical and X-ray afterglows, incorporating data obtained with BAT and XRT. The light curves show that the jet emission episode lasts $\sim 360$~s with a sharp pulse from BAT to XRT (Episode I). In Episode II, the optical emission has a smooth onset bump followed by a normal decay ($α_{\rm R,2} \approx -1.36$), as predicted in the standard external shock model, while the X-ray emission exhibits a plateau ($α_{\rm X,1} \approx -0.36$) followed by a steep decay ($α_{\rm X,2} \approx -2.12$). The light curves show obvious chromatic behavior with an excess in the X-ray flux. Our results suggest that GRB 150910A is an unusual GRB driven by a newly-born magnetar with its extremely energetic magnetic dipole (MD) wind in Episode II, which overwhelmingly dominates the observed early X-ray plateau. The radiative efficiency of the jet prompt emission is $η_γ \approx 11\%$. The MD wind emission was detected in both the BAT and XRT bands, making it the brightest among the current sample of MD winds seen by XRT. We infer the initial spin period ($P_0$) and the surface polar cap magnetic field strength ($B_p$) of the magnetar as $1.02 \times 10^{15}~{\rm G} \leq B_{p} \leq 1.80 \times 10^{15}~{\rm G}$ and 1~ms $\leq P_{0}v\leq 1.77$~ms, and the radiative efficiency of the wind is $η_w \geq 32\%$.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Magnetar as Central Engine of Gamma-Ray Bursts: Quasi-Universal Jet, Event Rate and X-ray Luminosity Function of Dipole Radiations

Early shallow-decaying X-ray afterglows of gamma-ray bursts (GRBs) may be attributed to the dipole radiations of newly-born magnetars. Assuming that the GRB jets powered by magnetars are quasi-universal, we find that the jet structure can be parameterized as a uniform jet with a luminosity of $\log L_{\rm j}/{\rm erg\ s^{-1}}=52.68^{+0.76}_{-0.33}$ (1$σ$) and an opening angle $θ_{\rm j}=2.10_{-1.28}^{+1.90}$ (50\% confidence level), surrounding by a power-law decay component with an index of ${-4.00^{+0.27}_{-0.37}}$ (1$σ$). The inferred local GRB rate is $ρ=9.6$ Gpc$^{-3}$ yr$^{-1}$ by including both the typical GRBs and LL-GRBs as the same population. The typical viewing angle is $3.3^{o}$, and may be $20^{o}\sim30^{o}$ for LL-GRBs. The X-ray luminosity function of the dipole radiation wind can be empirically described by a broken power-law function with indices $β_1=0.78^{+0.16}_{-0.15}$ and $β_2>1.6$ broken at $\log L_{b, w}/{\rm erg\ s^{-1}}=48.51^{+0.53}_{-0.65}$. In case of that the wind outflow is collimated and co-axial with the GRB jet, we find that the wind structure is similar to the GRB jet, i.e., $\log L_{\rm c, w}/{\rm erg\ s^{-1}}=48.38^{+0.30}_{-0.48}$, $θ_{\rm c, w}={2.65^{o}}_{-1.19^{o}}^{+0.1.73^{o}}$, and $k_{\rm w}=4.57^{+1.21}_{-0.75}$. The observed correlation between the prompt gamma-ray luminosity and X-ray luminosity of the wind may be resulted from the viewing angle effect in such a jet-wind system. Discussion on survey with the X-ray instruments on board the {\em Einstein\ Probe} mission in the soft X-ray band for the jet and wind emission is also presented.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Skeleton Based Action Recognition using a Stacked Denoising Autoencoder with Constraints of Privileged Information

Recently, with the availability of cost-effective depth cameras coupled with real-time skeleton estimation, the interest in skeleton-based human action recognition is renewed. Most of the existing skeletal representation approaches use either the joint location or the dynamics model. Differing from the previous studies, we propose a new method called Denoising Autoencoder with Temporal and Categorical Constraints (DAE_CTC)} to study the skeletal representation in a view of skeleton reconstruction. Based on the concept of learning under privileged information, we integrate action categories and temporal coordinates into a stacked denoising autoencoder in the training phase, to preserve category and temporal feature, while learning the hidden representation from a skeleton. Thus, we are able to improve the discriminative validity of the hidden representation. In order to mitigate the variation resulting from temporary misalignment, a new method of temporal registration, called Locally-Warped Sequence Registration (LWSR), is proposed for registering the sequences of inter- and intra-class actions. We finally represent the sequences using a Fourier Temporal Pyramid (FTP) representation and perform classification using a combination of LWSR registration, FTP representation, and a linear Support Vector Machine (SVM). The experimental results on three action data sets, namely MSR-Action3D, UTKinect-Action, and Florence3D-Action, show that our proposal performs better than many existing methods and comparably to the state of the art.

0 citations0 reviewsNo code linkedOpen access