arXiv subjects

Zhen Zhu

Publications and source records attributed to Zhen Zhu.

At least 55 records · Page 3Linked to original sources

arXiv2021

Semantic Flow for Fast and Accurate Scene Parsing

In this paper, we focus on designing effective method for fast and accurate scene parsing. A common practice to improve the performance is to attain high resolution feature maps with strong semantic representation. Two strategies are widely used -- atrous convolutions and feature pyramid fusion, are either computation intensive or ineffective. Inspired by the Optical Flow for motion alignment between adjacent video frames, we propose a Flow Alignment Module (FAM) to learn Semantic Flow between feature maps of adjacent levels, and broadcast high-level features to high resolution features effectively and efficiently. Furthermore, integrating our module to a common feature pyramid structure exhibits superior performance over other real-time methods even on light-weight backbone networks, such as ResNet-18. Extensive experiments are conducted on several challenging datasets, including Cityscapes, PASCAL Context, ADE20K and CamVid. Especially, our network is the first to achieve 80.4\% mIoU on Cityscapes with a frame rate of 26 FPS. The code is available at \url{https://github.com/lxtGH/SFSegNets}.

cs.CV↗

arXiv2021

Progressive and Aligned Pose Attention Transfer for Person Image Generation

This paper proposes a new generative adversarial network for pose transfer, i.e., transferring the pose of a given person to a target pose. We design a progressive generator which comprises a sequence of transfer blocks. Each block performs an intermediate transfer step by modeling the relationship between the condition and the target poses with attention mechanism. Two types of blocks are introduced, namely Pose-Attentional Transfer Block (PATB) and Aligned Pose-Attentional Transfer Bloc ~(APATB). Compared with previous works, our model generates more photorealistic person images that retain better appearance consistency and shape consistency compared with input images. We verify the efficacy of the model on the Market-1501 and DeepFashion datasets, using quantitative and qualitative measures. Furthermore, we show that our method can be used for data augmentation for the person re-identification task, alleviating the issue of data insufficiency. Code and pretrained models are available at https://github.com/tengteng95/Pose-Transfer.git.

cs.CV↗

arXiv2021

FaceController: Controllable Attribute Editing for Face in the Wild

Face attribute editing aims to generate faces with one or multiple desired face attributes manipulated while other details are preserved. Unlike prior works such as GAN inversion, which has an expensive reverse mapping process, we propose a simple feed-forward network to generate high-fidelity manipulated faces. By simply employing some existing and easy-obtainable prior information, our method can control, transfer, and edit diverse attributes of faces in the wild. The proposed method can consequently be applied to various applications such as face swapping, face relighting, and makeup transfer. In our method, we decouple identity, expression, pose, and illumination using 3D priors; separate texture and colors by using region-wise style codes. All the information is embedded into adversarial learning by our identity-style normalization module. Disentanglement losses are proposed to enhance the generator to extract information independently from each attribute. Comprehensive quantitative and qualitative evaluations have been conducted. In a single framework, our method achieves the best or competitive scores on a variety of face applications.

cs.CV↗

arXiv2021

Sustainable Border Control Policy in the COVID-19 Pandemic: A Math Modeling Study

Imported COVID-19 cases, if unchecked, can jeopardize the effort of domestic containment. We aim to find out what sustainable border control options for different entities (e.g., countries, states) exist during the reopening phases, given their own choice of domestic control measures and new technologies such as contact tracing. We propose a SUIHR model, which represents an extension to the discrete time SIR models. The model focuses on studying the spreading of virus predominantly by asymptomatic and pre-symptomatic patients. Imported risk and (1-tier) contact tracing are both built into the model. Under plausible parameter assumptions, we seek sustainable border control policies, in combination with sufficient internal measures, which allow entities to confine the virus without the need to revert back to more restrictive life styles or to rely on herd immunity. When the base reproduction number of COVID-19 exceeds 2.5, even 100% effective contact tracing alone is not enough to contain the spreading. For an entity that has completely eliminated the virus domestically, and resumes "normal", very strict pre-departure screening and test and isolation upon arrival combined with effective contact tracing can only delay another outbreak by 6 months. However, if the total net imported cases are non-increasing, and the entity employs a confining domestic control policy, then the total new cases can be contained even without border control.

q-bio.PE↗

arXiv2020

WDNet: Watermark-Decomposition Network for Visible Watermark Removal

Visible watermarks are widely-used in images to protect copyright ownership. Analyzing watermark removal helps to reinforce the anti-attack techniques in an adversarial way. Current removal methods normally leverage image-to-image translation techniques. Nevertheless, the uncertainty of the size, shape, color and transparency of the watermarks set a huge barrier for these methods. To combat this, we combine traditional watermarked image decomposition into a two-stage generator, called Watermark-Decomposition Network (WDNet), where the first stage predicts a rough decomposition from the whole watermarked image and the second stage specifically centers on the watermarked area to refine the removal results. The decomposition formulation enables WDNet to separate watermarks from the images rather than simply removing them. We further show that these separated watermarks can serve as extra nutrients for building a larger training dataset and further improving removal performance. Besides, we construct a large-scale dataset named CLWD, which mainly contains colored watermarks, to fill the vacuum of colored watermark removal dataset. Extensive experiments on the public gray-scale dataset LVW and CLWD consistently show that the proposed WDNet outperforms the state-of-the-art approaches both in accuracy and efficiency. The code and CLWD dataset are publicly available at https://github.com/MRUIL/WDNet.

cs.CV↗

arXiv2020

Discovery of segmented Fermi surface induced by Cooper pair momentum

Since the early days of Bardeen-Cooper-Schrieffer theory, it has been predicted that a sufficiently large supercurrent can close the energy gap in a superconductor and creates gapless Bogoliubov quasiparticles through the Doppler shift of quasiparticle energy due to the Cooper pair momentum. In this gapless superconducting state, zero-energy quasiparticles reside on a segment of the normal state Fermi surface, while its remaining part is still gapped. The finite density of states of field-induced quasiparticles, known as the Volovik effect, has been observed in tunneling and specific heat measurements on d- and s-wave superconductors. However, the segmented Fermi surface of a finite-momentum state carrying a supercurrent has never been detected directly. Here we use quasiparticle interference (QPI) technique to image field-controlled Fermi surface of Bi$_2$Te$_3$ thin films proximitized by the superconductor NbSe$_2$. By applying a small in-plane magnetic field, a screening supercurrent is induced which leads to finite-momentum pairing on topological surface states of Bi$_2$Te$_3$. Our measurements and analysis reveal the strong impact of finite Cooper pair momentum on the quasiparticle spectrum, and thus pave the way for STM study of pair density wave and FFLO states in unconventional superconductors.

cond-mat.supr-con↗

arXiv2020

Counterfactual Learning to Rank using Heterogeneous Treatment Effect Estimation

Learning-to-Rank (LTR) models trained from implicit feedback (e.g. clicks) suffer from inherent biases. A well-known one is the position bias -- documents in top positions are more likely to receive clicks due in part to their position advantages. To unbiasedly learn to rank, existing counterfactual frameworks first estimate the propensity (probability) of missing clicks with intervention data from a small portion of search traffic, and then use inverse propensity score (IPS) to debias LTR algorithms on the whole data set. These approaches often assume the propensity only depends on the position of the document, which may cause high estimation variance in applications where the search context (e.g. query, user) varies frequently. While context-dependent propensity models reduce variance, accurate estimations may require randomization or intervention on a large amount of traffic, which may not be realistic in real-world systems, especially for long tail queries. In this work, we employ heterogeneous treatment effect estimation techniques to estimate position bias when intervention click data is limited. We then use such estimations to debias the observed click distribution and re-draw a new de-biased data set, which can be used for any LTR algorithms. We conduct simulations with varying experiment conditions and show the effectiveness of the proposed method in regimes with long tail queries and sparse clicks.

cs.IR↗

arXiv2020

Semantically Multi-modal Image Synthesis

In this paper, we focus on semantically multi-modal image synthesis (SMIS) task, namely, generating multi-modal images at the semantic level. Previous work seeks to use multiple class-specific generators, constraining its usage in datasets with a small number of classes. We instead propose a novel Group Decreasing Network (GroupDNet) that leverages group convolutions in the generator and progressively decreases the group numbers of the convolutions in the decoder. Consequently, GroupDNet is armed with much more controllability on translating semantic labels to natural images and has plausible high-quality yields for datasets with many classes. Experiments on several challenging datasets demonstrate the superiority of GroupDNet on performing the SMIS task. We also show that GroupDNet is capable of performing a wide range of interesting synthesis applications. Codes and models are available at: https://github.com/Seanseattle/SMIS.

cs.CV↗

arXiv2019

Highly Mobile Carriers in a Candidate of Quasi-Two-Dimensional Topological Semimetal AuTe$_2$Br

We report the crystal and electronic structures of a non-centrosymmetric quasi-two-dimensional (2D), candidate of topological semimetal AuTe2Br. The Fermi surface of this layered compound consists of 2D-like, topological trivial electron and non-trivial hole pockets which host a Dirac cone along the kz direction. Our transport measurements on the single crystals show highly anisotropic, compensated low-density electrons and holes, both of which exhibit ultrahigh mobility at a level of 10^5cm^2V^-1s^-1 at low temperature. The highly mobile, compensated carriers lead a non-saturated, parabolic magnetoresistance as large as 3*10^5 in single-crystalline AuTe2Br in a magnetic field up to 58 T.

cond-mat.mtrl-sci↗

arXiv2019

Asymmetric Non-local Neural Networks for Semantic Segmentation

The non-local module works as a particularly useful technique for semantic segmentation while criticized for its prohibitive computation and GPU memory occupation. In this paper, we present Asymmetric Non-local Neural Network to semantic segmentation, which has two prominent components: Asymmetric Pyramid Non-local Block (APNB) and Asymmetric Fusion Non-local Block (AFNB). APNB leverages a pyramid sampling module into the non-local block to largely reduce the computation and memory consumption without sacrificing the performance. AFNB is adapted from APNB to fuse the features of different levels under a sufficient consideration of long range dependencies and thus considerably improves the performance. Extensive experiments on semantic segmentation benchmarks demonstrate the effectiveness and efficiency of our work. In particular, we report the state-of-the-art performance of 81.3 mIoU on the Cityscapes test set. For a 256x128 input, APNB is around 6 times faster than a non-local block on GPU while 28 times smaller in GPU running memory occupation. Code is available at: https://github.com/MendelXu/ANN.git.

cs.CV↗

arXiv2019

The Unusual Suppression of Superconducting Transition Temperature in Double-Doping 2H-NbSe$_2$

2H-NbSe2 is one of the most widely researched transition metal dichalcogenide (TMD) superconductors, which undergoes charge-density wave (CDW) transition at TCDW about 33 K and superconducting transition at Tc of 7.3 K. To explore the relation between its superconductivity and Fermi surface nesting, we combined S substitution with Cu intercalation in 2H-NbSe2 to make CuxNbSe2-ySy. Upon systematic substitution of S and intercalation of Cu ions into 2H-NbSe2, we found that when the Cu and S contents increases, the Tc decreases in CuxNbSe2-ySy. While at higher x and y values, Tc keeps a constant value near 2 K, which is not commonly observed for a layered TMD. For comparison, we found the simultaneous substitution of Nb by Cu and Se by S in CuxNb1-xSe2-ySy lowered the Tc substantially faster. We construct a superconducting phase diagrams for our double-doping compounds in contrast with the related single-ions doping systems.

cond-mat.supr-con↗

arXiv2019

DOTA: A Large-scale Dataset for Object Detection in Aerial Images

Object detection is an important and challenging problem in computer vision. Although the past decade has witnessed major advances in object detection in natural scenes, such successes have been slow to aerial imagery, not only because of the huge variation in the scale, orientation and shape of the object instances on the earth's surface, but also due to the scarcity of well-annotated datasets of objects in aerial scenes. To advance object detection research in Earth Vision, also known as Earth Observation and Remote Sensing, we introduce a large-scale Dataset for Object deTection in Aerial images (DOTA). To this end, we collect $2806$ aerial images from different sensors and platforms. Each image is of the size about 4000-by-4000 pixels and contains objects exhibiting a wide variety of scales, orientations, and shapes. These DOTA images are then annotated by experts in aerial image interpretation using $15$ common object categories. The fully annotated DOTA images contains $188,282$ instances, each of which is labeled by an arbitrary (8 d.o.f.) quadrilateral To build a baseline for object detection in Earth Vision, we evaluate state-of-the-art object detection algorithms on DOTA. Experiments demonstrate that DOTA well represents real Earth Vision applications and are quite challenging.

cs.CV↗

arXiv2019

Progressive Pose Attention Transfer for Person Image Generation

This paper proposes a new generative adversarial network for pose transfer, i.e., transferring the pose of a given person to a target pose. The generator of the network comprises a sequence of Pose-Attentional Transfer Blocks that each transfers certain regions it attends to, generating the person image progressively. Compared with those in previous works, our generated person images possess better appearance consistency and shape consistency with the input images, thus significantly more realistic-looking. The efficacy and efficiency of the proposed network are validated both qualitatively and quantitatively on Market-1501 and DeepFashion. Furthermore, the proposed architecture can generate training images for person re-identification, alleviating data insufficiency. Codes and models are available at: https://github.com/tengteng95/Pose-Transfer.git.

cs.CV↗

arXiv2019

Observation of ballistic avalanche phenomena in nanoscale vertical InSe/BP heterostructures

Initiating impact ionization of avalanche breakdown essentially requires applying a high electric field in a long active region, hampering carrier-multiplication with high gain, low bias and superior noise performance. Here we report the observation of ballistic avalanche phenomena in sub-MFP scaled vertical indium selenide (InSe)/black phosphorus (BP) heterostructures. The heterojunction is engineered to avalanche photodetectors (APD) and impact ionization transistors, demonstrating ultra-sensitive mid-IR light detection (4 μm wavelength) and ultra-steep subthreshold swing, respectively. These devices show an extremely low avalanche threshold (<1 volt), excellent low noise figures and distinctive density spectral shape. Further transport measurement evidences the breakdown originals from a ballistic avalanche phenomenon, where the sub-MFP BP channel enables both electrons and holes to impact-ionize the lattice and abruptly amplify the current without scattering from the obstacles in a deterministic nature. Our results shed light on the development of advanced photodetectors and efficiently facilitating carriers on the nanoscale.

cond-mat.mtrl-sci↗

arXiv2018

Can the fluctuations of the quantum vacuum solve the cosmological constant problem?

The cosmological constant problem arises because the magnitude of vacuum energy density predicted by quantum mechanics is about 120 orders of magnitude larger than the value implied by cosmological observations of accelerating cosmic expansion. Recently, some of the current authors proposed that the stochastic nature of the quantum vacuum can resolve this tension [Q, Wang, Z. Zhu, and W. G. Unruh, Phys. Rev. D 95, 103504, 2017]. By treating the fluctuations in the vacuum seriously and allowing fluctuations up to some high-energy cutoff at which Quantum Field Theory is believed to break down, a parametric resonance effect arises that leads to a slow expansion and acceleration. In this work, we thoroughly examine the implications of this proposal by investigating the resulting dynamics. First, we improve upon numerical calculations in the original work and show that convergence issues had overshadowed some important effects. Correct calculations reverse some of the conclusions in [Q. Wang, Z. Zhu, and W. G. Unruh, Phys. Rev. D 95, 103504, 2017], however the premise that parametric resonance can explain a very slowly accelerating expansion appears to remain sound. After improving the resolution and efficiency of the numerical tests, we explore a wider range of cutoff energies, and examine the effects of multiple particle fields. We introduce a simple model using the Mathieu equation (a prototypical example of parametric resonance), and find that it closely matches numerical results in regimes where its assumptions are valid. Using this model, we extrapolate to find that in a universe with 28 bosonic fields and a high-energy cutoff 40 times higher than the Planck energy, the acceleration would be comparable to what is observed.

gr-qc↗

arXiv2018

Hierarchical Modeling and Shrinkage for User Session Length Prediction in Media Streaming

An important metric of users' satisfaction and engagement within on-line streaming services is the user session length, i.e. the amount of time they spend on a service continuously without interruption. Being able to predict this value directly benefits the recommendation and ad pacing contexts in music and video streaming services. Recent research has shown that predicting the exact amount of time spent is highly nontrivial due to many external factors for which a user can end a session, and the lack of predictive covariates. Most of the other related literature on duration based user engagement has focused on dwell time for websites, for search and display ads, mainly for post-click satisfaction prediction or ad ranking. In this work we present a novel framework inspired by hierarchical Bayesian modeling to predict, at the moment of login, the amount of time a user will spend in the streaming service. The time spent by a user on a platform depends upon user-specific latent variables which are learned via hierarchical shrinkage. Our framework enjoys theoretical guarantees and naturally incorporates flexible parametric/nonparametric models on the covariates, including models robust to outliers. Our proposal is found to outperform state-of- the-art estimators in terms of efficiency and predictive performance on real world public and private datasets.

stat.ML↗

arXiv2018

Non-Stationary Texture Synthesis by Adversarial Expansion

The real world exhibits an abundance of non-stationary textures. Examples include textures with large-scale structures, as well as spatially variant and inhomogeneous textures. While existing example-based texture synthesis methods can cope well with stationary textures, non-stationary textures still pose a considerable challenge, which remains unresolved. In this paper, we propose a new approach for example-based non-stationary texture synthesis. Our approach uses a generative adversarial network (GAN), trained to double the spatial extent of texture blocks extracted from a specific texture exemplar. Once trained, the fully convolutional generator is able to expand the size of the entire exemplar, as well as of any of its sub-blocks. We demonstrate that this conceptually simple approach is highly effective for capturing large-scale structures, as well as other non-stationary attributes of the input exemplar. As a result, it can cope with challenging textures, which, to our knowledge, no other existing method can handle.

cs.GR↗

arXiv2018

Rotation-Sensitive Regression for Oriented Scene Text Detection

Text in natural images is of arbitrary orientations, requiring detection in terms of oriented bounding boxes. Normally, a multi-oriented text detector often involves two key tasks: 1) text presence detection, which is a classification problem disregarding text orientation; 2) oriented bounding box regression, which concerns about text orientation. Previous methods rely on shared features for both tasks, resulting in degraded performance due to the incompatibility of the two tasks. To address this issue, we propose to perform classification and regression on features of different characteristics, extracted by two network branches of different designs. Concretely, the regression branch extracts rotation-sensitive features by actively rotating the convolutional filters, while the classification branch extracts rotation-invariant features by pooling the rotation-sensitive features. The proposed method named Rotation-sensitive Regression Detector (RRD) achieves state-of-the-art performance on three oriented scene text benchmark datasets, including ICDAR 2015, MSRA-TD500, RCTW-17 and COCO-Text. Furthermore, RRD achieves a significant improvement on a ship collection dataset, demonstrating its generality on oriented object detection.

cs.CV↗