Search arXiv⌕ Search

arXiv subjects

Zheng Zhu

Publications and source records attributed to Zheng Zhu.

At least 253 records · Page 14Linked to original sources
arXiv2019

Exploiting Offset-guided Network for Pose Estimation and Tracking

Human pose estimation has witnessed a significant advance thanks to the development of deep learning. Recent human pose estimation approaches tend to directly predict the location heatmaps, which causes quantization errors and inevitably deteriorates the performance within the reduced network output. Aim at solving it, we revisit the heatmap-offset aggregation method and propose the Offset-guided Network (OGN) with an intuitive but effective fusion strategy for both two-stages pose estimation and Mask R-CNN. For two-stages pose estimation, a greedy box generation strategy is also proposed to keep more necessary candidates while performing person detection. For mask R-CNN, ratio-consistent is adopted to improve the generalization ability of the network. State-of-the-art results on COCO and PoseTrack dataset verify the effectiveness of our offset-guided pose estimation and tracking.

cs.CV↗
arXiv2019

State-aware Re-identification Feature for Multi-target Multi-camera Tracking

Multi-target Multi-camera Tracking (MTMCT) aims to extract the trajectories from videos captured by a set of cameras. Recently, the tracking performance of MTMCT is significantly enhanced with the employment of re-identification (Re-ID) model. However, the appearance feature usually becomes unreliable due to the occlusion and orientation variance of the targets. Directly applying Re-ID model in MTMCT will encounter the problem of identity switches (IDS) and tracklet fragment caused by occlusion. To solve these problems, we propose a novel tracking framework in this paper. In this framework, the occlusion status and orientation information are utilized in Re-ID model with human pose information considered. In addition, the tracklet association using the proposed fused tracking feature is adopted to handle the fragment problem. The proposed tracker achieves 81.3\% IDF1 on the multiple-camera hard sequence, which outperforms all other reference methods by a large margin.

cs.CV↗
arXiv2019

Attention-guided Unified Network for Panoptic Segmentation

This paper studies panoptic segmentation, a recently proposed task which segments foreground (FG) objects at the instance level as well as background (BG) contents at the semantic level. Existing methods mostly dealt with these two problems separately, but in this paper, we reveal the underlying relationship between them, in particular, FG objects provide complementary cues to assist BG understanding. Our approach, named the Attention-guided Unified Network (AUNet), is a unified framework with two branches for FG and BG segmentation simultaneously. Two sources of attentions are added to the BG branch, namely, RPN and FG segmentation mask to provide object-level and pixel-level attentions, respectively. Our approach is generalized to different backbones with consistent accuracy gain in both FG and BG segmentation, and also sets new state-of-the-arts both in the MS-COCO (46.5% PQ) and Cityscapes (59.0% PQ) benchmarks.

cs.CV↗
arXiv2019

Deformations of Bi-conformal Energy and a new Characterization of Quasiconformality

The concept of hyperelastic deformations of bi-conformal energy is developed as an extension of quasiconformality. These are homeomorphisms $h:X \to Y$ between domains $ X, Y \subset \mathbb R^n$ of the Sobolev class $W^{1,n}_{loc} (X, Y)$ whose inverse $f =h^{-1}:Y \to X$ also belongs to $W^{1,n}_{loc}(Y, X)$. Thus the paper opens new topics in Geometric Function Theory with connections to mathematical models of Nonlinear Elasticity. In seeking differences and similarities with quasiconformal mappings we examine closely the modulus of continuity of deformations of bi-conformal energy. This leads us to a new characterization of quasiconformality. Specifically, it is observed that quasiconformal mappings behave locally at every point like radial stretchings. Without going into detail, if a quasiconformal map $h$ admits a function $ϕ$ as its optimal modulus of continuity at a point $x_0$, then $f = h^{-1}$ admits the inverse function $ψ= ϕ^{-1}$ as its modulus of continuity at $y_0 = h(x_0)$. That is to say; a poor continuity of $h$ at a given point $x_0$ is always compensated by a better continuity of $f$ at $y_0$, and vice versa. Such a gain/loss property, seemingly overlooked by many authors, is actually characteristic of quasiconformal mappings. It turns out that the elastic deformations of bi-conformal energy are very different in this respect. Unexpectedly, such a map may have the same optimal modulus of continuity as its inverse deformation. In line with Hooke's Law, when trying to restore the original shape of the body (by the inverse transformation) the modulus of continuity may neither be improved nor become worse. However, examples to confirm this phenomenon are far from being obvious. We eventually hope that our examples will gain an interest in the materials science, particularly in mathematical models of hyperelasticity.

math.CA↗
arXiv2019

Do tensor renormalization group methods work for frustrated spin systems?

No. To illustrate that tensor renormalization group methods are poorly suited for frustrated magnetic systems, we study the thermodynamic properties of the two-dimensional Edwards-Anderson Ising spin-glass model on a square lattice. We show that the limited precision of standard 64-bit data types and not a small cut-off parameter is the main reason for unphysical negative partition function values in spin glasses.

cond-mat.dis-nn↗
arXiv2019

Motion Control on Bionic Eyes: A Comprehensive Review

Biology can provide biomimetic components and new control principles for robotics. Developing a robot system equipped with bionic eyes is a difficult but exciting task. Researchers have been studying the control mechanisms of bionic eyes for many years and considerable models are available. In this paper, control model and its implementation on robots for bionic eyes are reviewed, which covers saccade, smooth pursuit, vergence, vestibule-ocular reflex (VOR), optokinetic reflex (OKR) and eye-head coordination. What is more, some problems and possible solutions in the field of bionic eyes are discussed and analyzed. This review paper can be used as a guide for researchers to identify potential research problems and solutions of the bionic eyes' motion control.

cs.RO↗
arXiv2018

Action Machine: Rethinking Action Recognition in Trimmed Videos

Existing methods in video action recognition mostly do not distinguish human body from the environment and easily overfit the scenes and objects. In this work, we present a conceptually simple, general and high-performance framework for action recognition in trimmed videos, aiming at person-centric modeling. The method, called Action Machine, takes as inputs the videos cropped by person bounding boxes. It extends the Inflated 3D ConvNet (I3D) by adding a branch for human pose estimation and a 2D CNN for pose-based action recognition, being fast to train and test. Action Machine can benefit from the multi-task training of action recognition and pose estimation, the fusion of predictions from RGB images and poses. On NTU RGB-D, Action Machine achieves the state-of-the-art performance with top-1 accuracies of 97.2% and 94.3% on cross-view and cross-subject respectively. Action Machine also achieves competitive performance on another three smaller action recognition datasets: Northwestern UCLA Multiview Action3D, MSR Daily Activity3D and UTD-MHAD. Code will be made available.

cs.CV↗
arXiv2018

Two-hole ground state wavefunction: Non-BCS pairing in a $t$-$J$ two-leg ladder system

Superconductivity is usually described in the framework of the Bardeen-Cooper-Schrieffer (BCS) wavefunction, which even includes the resonating-valence-bond (RVB) wavefunction proposed for the high-temperature superconductivity in the cuprate. A natural question is \emph{if} any fundamental physics could be possibly missed by applying such a scheme to strongly correlated systems. Here we study the pairing wavefunction of two holes injected into a Mott insulator/antiferromagnet in a two-leg ladder using variational Monte Carlo (VMC) approach. By comparing with density matrix renormalization group (DMRG) calculation, we show that a conventional BCS or RVB pairing of the doped holes makes qualitatively wrong predictions and is incompatible with the fundamental pairing force in the $t$-$J$ model, which is kinetic-energy-driven by nature. By contrast, a non-BCS-like wavefunction incorporating such novel effect will result in a substantially enhanced pairing strength and improved ground state energy as compared to the DMRG results. We argue that the non-BCS form of such a new ground state wavefunction is essential to describe a doped Mott antiferromagnet at finite doping.

cond-mat.str-el↗
arXiv2018

Targeted Learning with Daily EHR Data

Electronic health records (EHR) data provide a cost and time-effective opportunity to conduct cohort studies of the effects of multiple time-point interventions in the diverse patient population found in real-world clinical settings. Because the computational cost of analyzing EHR data at daily (or more granular) scale can be quite high, a pragmatic approach has been to partition the follow-up into coarser intervals of pre-specified length. Current guidelines suggest employing a 'small' interval, but the feasibility and practical impact of this recommendation has not been evaluated and no formal methodology to inform this choice has been developed. We start filling these gaps by leveraging large-scale EHR data from a diabetes study to develop and illustrate a fast and scalable targeted learning approach that allows to follow the current recommendation and study its practical impact on inference. More specifically, we map daily EHR data into four analytic datasets using 90, 30, 15 and 5-day intervals. We apply a semi-parametric and doubly robust estimation approach, the longitudinal TMLE, to estimate the causal effects of four dynamic treatment rules with each dataset, and compare the resulting inferences. To overcome the computational challenges presented by the size of these data, we propose a novel TMLE implementation, the 'long-format TMLE', and rely on the latest advances in scalable data-adaptive machine-learning software, xgboost and h2o, for estimation of the TMLE nuisance parameters.

stat.AP↗
arXiv2018

Identity-Enhanced Network for Facial Expression Recognition

Facial expression recognition is a challenging task, arguably because of large intra-class variations and high inter-class similarities. The core drawback of the existing approaches is the lack of ability to discriminate the changes in appearance caused by emotions and identities. In this paper, we present a novel identity-enhanced network (IDEnNet) to eliminate the negative impact of identity factor and focus on recognizing facial expressions. Spatial fusion combined with self-constrained multi-task learning are adopted to jointly learn the expression representations and identity-related information. We evaluate our approach on three popular datasets, namely Oulu-CASIA, CK+ and MMI. IDEnNet improves the baseline consistently, and achieves the best or comparable state-of-the-art on all three datasets.

cs.CV↗
arXiv2018

Multi-hierarchical Independent Correlation Filters for Visual Tracking

For visual tracking, most of the traditional correlation filters (CF) based methods suffer from the bottleneck of feature redundancy and lack of motion information. In this paper, we design a novel tracking framework, called multi-hierarchical independent correlation filters (MHIT). The framework consists of motion estimation module, hierarchical features selection, independent CF online learning, and adaptive multi-branch CF fusion. Specifically, the motion estimation module is introduced to capture motion information, which effectively alleviates the object partial occlusion in the temporal video. The multi-hierarchical deep features of CNN representing different semantic information can be fully excavated to track multi-scale objects. To better overcome the deep feature redundancy, each hierarchical features are independently fed into a single branch to implement the online learning of parameters. Finally, an adaptive weight scheme is integrated into the framework to fuse these independent multi-branch CFs for the better and more robust visual object tracking. Extensive experiments on OTB and VOT datasets show that the proposed MHIT tracker can significantly improve the tracking performance. Especially, it obtains a 20.1% relative performance gain compared to the top trackers on the VOT2017 challenge, and also achieves new state-of-the-art performance on the VOT2018 challenge.

cs.CV↗
arXiv2018

An Efficient Optical Flow Based Motion Detection Method for Non-stationary Scenes

Real-time motion detection in non-stationary scenes is a difficult task due to dynamic background, changing foreground appearance and limited computational resource. These challenges degrade the performance of the existing methods in practical applications. In this paper, an optical flow based framework is proposed to address this problem. By applying a novel strategy to utilize optical flow, we enable our method being free of model constructing, training or updating and can be performed efficiently. Besides, a dual judgment mechanism with adaptive intervals and adaptive thresholds is designed to heighten the system's adaptation to different situations. In experiment part, we quantitatively and qualitatively validate the effectiveness and feasibility of our method with videos in various scene conditions. The experimental results show that our method adapts itself to different situations and outperforms the state-of-the-art real-time methods, indicating the advantages of our optical flow based method.

cs.CV↗
arXiv2018

Meson formation in mixed-dimensional t-J models

Surprising properties of doped Mott insulators are at the heart of many quantum materials, including transition metal oxides and organic materials. The key to unraveling complex phenomena observed in these systems lies in understanding the interplay of spin and charge degrees of freedom. One of the most debated questions concerns the nature of charge carriers in a background of fluctuating spins. To shed new light on this problem, we suggest a simplified model with mixed dimensionality, where holes move through a Mott insulator unidirectionally while spin exchange interactions are two dimensional. By studying individual holes in this system, we find direct evidence for the formation of mesonic bound states of holons and spinons, connected by a string of displaced spins -- a precursor of the spin-charge separation obtained in the 1D limit of the model. Our predictions can be tested using ultracold atoms in a quantum gas microscope, allowing to directly image spinons and holons, and reveal the short-range hidden string order which we predict in this model.

cond-mat.str-el↗
arXiv2018

Optical Flow Based Online Moving Foreground Analysis

Obtained by moving object detection, the foreground mask result is unshaped and can not be directly used in most subsequent processes. In this paper, we focus on this problem and address it by constructing an optical flow based moving foreground analysis framework. During the processing procedure, the foreground masks are analyzed and segmented through two complementary clustering algorithms. As a result, we obtain the instance-level information like the number, location and size of moving objects. The experimental result show that our method adapts itself to the problem and performs well enough for practical applications.

cs.CV↗
arXiv2018

Localization in a $t$-$J$ type ladder with translational symmetry

An explicit $\textit{spatial}$ localization of a hole is shown in a two-leg $t$-$J$ ladder in the presence of a staggered chemical potential, which still retains a translational symmetry, by density matrix renormalization group method. Delocalization can be recovered in the following cases, where either the hidden phase string effect is turned off or a finite next-nearest-neighbor hopping $t'$ is added to sufficiently weaken the phase string effect. In addition, two holes are always delocalized by forming a mobile bound pair, in contrast to the localized single holes, which points to a novel pairing mechanism as one of the essential properties of a doped Mott insulator.

cond-mat.str-el↗
arXiv2018

Valley Stoner Instability of the Composite Fermi Sea

We study two-component electrons in the lowest Landau level at total filling factor $ν_T=1/2$ with anisotropic mass tensors and principal axes rotated by $π/2$ as realized in Aluminum Arsenide (AlAs) quantum wells. Combining exact diagonalization and the density matrix renormalization group we demonstrate that the system undergoes a quantum phase transition from a gapless state in which both flavors are equally populated to another gapless state in which all the electrons spontaneously polarize into a single flavor beyond a critical mass anisotropy of {\bf $m_x/m_y \sim 7$}. We propose that this phase transition is a form of itinerant Stoner transition between a two-component and a single-component composite fermi sea states and describe a set of trial wavefunctions which successfully capture the quantum numbers and shell filling effects in finite size systems as well as providing a physical picture for the energetics of these states. Our estimates indicate that the composite Fermi sea of AlAs is the analog of an itinerant Stoner magnet with a finite spontaneous valley polarization. We pinpoint experimental evidence indicating the presence of Stoner magnetism in the Jain states surrounding $ν=1/2$.

cond-mat.str-el↗
arXiv2018

Exciton Condensation in Quantum Hall Bilayers at Total Filling $ν_T=5$

We study the coupled quantum Hall bilayers each at half-filled first excited Landau levels with varying the layer distance. Based on numerical exact diagonalization on torus, we identify two distinct phases separated by a critical layer distance $d_c$. From $d_c$ to infinite layer distance, the topological phase is smoothly connected to a direct tensor product of two Moore-Read states, while the interlayer coherence emerges at $d<d_c$ characterized by the $xy$ easy-plane ferromagnetic energy spectra, gapless pseudospin excitations and the finite exciton superfluid stiffness, corresponding to the exciton superfluid state. More interestingly, the results of the ground state fidelity, the evolution of energy spectra, and the superfluid stiffness indicate a possible continuous transition. Theoretically it can be interpreted as a topological phase transition which simultaneously changes the topology of ground state and breaks symmetry, providing an interesting example of transitions beyond Landau paradigm.

cond-mat.str-el↗