arXiv subjects

Tao Wang

Publications and source records attributed to Tao Wang.

At least 739 records · Page 41Linked to original sources

arXiv2020

A multiple attributes image quality database for smartphone camera photo quality assessment

Smartphone is the superstar product in digital device market and the quality of smartphone camera photos (SCPs) is becoming one of the dominant considerations when consumers purchase smartphones. How to evaluate the quality of smartphone cameras and the taken photos is urgent issue to be solved. To bridge the gap between academic research accomplishment and industrial needs, in this paper, we establish a new Smartphone Camera Photo Quality Database (SCPQD2020) including 1800 images with 120 scenes taken by 15 smartphones. Exposure, color, noise and texture which are four dominant factors influencing the quality of SCP are evaluated in the subjective study, respectively. Ten popular no-reference (NR) image quality assessment (IQA) algorithms are tested and analyzed on our database. Experimental results demonstrate that the current objective models are not suitable for SCPs, and quality metrics having high correlation with human visual perception are highly needed.

eess.IV↗

arXiv2020

InSe: a two-dimensional semiconductor with superior flexibility

Two-dimensional Indium Selenide (InSe) has attracted extensive attention recently due to its record-high charge carrier mobility and photoresponsivity in the fields of electronics and optoelectronics. Nevertheless, the mechanical properties of this material in the ultra-thin regime have not been investigated yet. Here, we present our efforts to determine the Young's modulus of thin InSe (~1-2 layers to ~40 layers) flakes experimentally by using buckling-based methodology. We find that the Young's modulus has a value of 23.1 +- 5.2 GPa, one of the lowest values reported up to date for crystalline two-dimensional materials. This superior flexibility can be very attractive for different applications, such as strain engineering and flexible electronics.

physics.app-ph↗

arXiv2020

A Tamm Plasmon-Porous GaN Distributed Bragg Reflector Cavity

This paper reports on design, measurement and optimisation of a Tamm plasmon metal-DBR cavity for use in the green part of the visible spectrum. It uses an optimised silver layer thickness and a porous DBR created using a novel electro-chemical etching technique. This device has applications in low cost lasers, photodetectors and photoconductive switches for the visible wavelength range.

physics.optics↗

arXiv2020

Variational Prototype Replays for Continual Learning

Continual learning refers to the ability to acquire and transfer knowledge without catastrophically forgetting what was previously learned. In this work, we consider \emph{few-shot} continual learning in classification tasks, and we propose a novel method, Variational Prototype Replays, that efficiently consolidates and recalls previous knowledge to avoid catastrophic forgetting. In each classification task, our method learns a set of variational prototypes with their means and variances, where embedding of the samples from the same class can be represented in a prototypical distribution and class-representative prototypes are separated apart. To alleviate catastrophic forgetting, our method replays one sample per class from previous tasks, and correspondingly matches newly predicted embeddings to their nearest class-representative prototypes stored from previous tasks. Compared with recent continual learning approaches, our method can readily adapt to new tasks with more classes without requiring the addition of new units. Furthermore, our method is more memory efficient since only class-representative prototypes with their means and variances, as well as only one sample per class from previous tasks need to be stored. Without tampering with the performance on initial tasks, our method learns novel concepts given a few training examples of each class in new tasks.

cs.CV↗

arXiv2020

An inexpensive system for the deterministic transfer of 2D materials

The development of systems for the deterministic transfer of two-dimensional (2D) materials have undoubtedly contributed to a great advance in the 2D materials research. In fact, they have made it possible to fabricate van der Waals heterostructures and 2D materials-based devices with complex architectures. Nonetheless, as far as we know, the amount of papers in the literature providing enough details to reproduce these systems by other research groups is very scarce in the literature. Moreover, these systems typically require the use of expensive optical and mechanical components hampering their applicability in research groups with low budget. Here we demonstrate how a deterministic placement system for 2D materials set up with full capabilities can be implemented under 900 Eur which can be easily implemented in low budget labs and educational labs.

physics.ed-ph↗

arXiv2020

Growth, charge and thermal transport of flowered graphene

We report on the structural and transport properties of the smallest dislocation loop in graphene, known as a flower defect. First, by means of advanced experimental imaging techniques, we deduce how flower defects are formed during recrystallization of chemical vapor deposited graphene. We propose that the flower defects arise from a bulge type mechanism in which the flower domains are the grains left over by dynamic recrystallisation. Next, in order to evaluate the use of such defects as possible building blocks for all-graphene electronics, we combine multiscale modeling tools to investigate the structure and the electron and phonon transport properties of large monolayer graphene samples with a random distribution of flower defects. For large enough flower densities, we find that electron transport is strongly suppressed while, surprisingly, hole transport remains almost unaffected. These results suggest possible applications of flowered graphene for electron energy filtering. For the same defect densities, phonon transport is reduced by orders of magnitude as elastic scattering by defects becomes dominant. Heat transport by flexural phonons, key in graphene, is largely suppressed even for very low concentrations.

cond-mat.mtrl-sci↗

arXiv2020

CTM: Collaborative Temporal Modeling for Action Recognition

With the rapid development of digital multimedia, video understanding has become an important field. For action recognition, temporal dimension plays an important role, and this is quite different from image recognition. In order to learn powerful feature of videos, we propose a Collaborative Temporal Modeling (CTM) block (Figure 1) to learn temporal information for action recognition. Besides a parameter-free identity shortcut, as a separate temporal modeling block, CTM includes two collaborative paths: a spatial-aware temporal modeling path, which we propose the Temporal-Channel Convolution Module (TCCM) with unshared parameters for each spatial position (H*W) to build, and a spatial-unaware temporal modeling path. CTM blocks can seamlessly be inserted into many popular networks to generate CTM Networks and bring the capability of learning temporal information to 2D CNN backbone networks, which only capture spatial information. Experiments on several popular action recognition datasets demonstrate that CTM blocks bring the performance improvements on 2D CNN baselines, and our method achieves the competitive results against the state-of-the-art methods. Code will be made publicly available.

cs.CV↗

arXiv2020

iqiyi Submission to ActivityNet Challenge 2019 Kinetics-700 challenge: Hierarchical Group-wise Attention

In this report, the method for the iqiyi submission to the task of ActivityNet 2019 Kinetics-700 challenge is described. Three models are involved in the model ensemble stage: TSN, HG-NL and StNet. We propose the hierarchical group-wise non-local (HG-NL) module for frame-level features aggregation for video classification. The standard non-local (NL) module is effective in aggregating frame-level features on the task of video classification but presents low parameters efficiency and high computational cost. The HG-NL method involves a hierarchical group-wise structure and generates multiple attention maps to enhance performance. Basing on this hierarchical group-wise structure, the proposed method has competitive accuracy, fewer parameters and smaller computational cost than the standard NL. For the task of ActivityNet 2019 Kinetics-700 challenge, after model ensemble, we finally obtain an averaged top-1 and top-5 error percentage 28.444% on the test set.

cs.CV↗

arXiv2019

Analysis of the L1 scheme for fractional wave equations with nonsmooth data

This paper analyzes the well-known L1 scheme for fractional wave equations with nonsmooth data. A new stability estimate is obtained, and the temporal accuracy $ \mathcal O(τ^{3-α}) $ is derived for the nonsmooth initial data. In addition, a modified L1 scheme is proposed, and stability and temporal accuracy $ \mathcal O(τ^2) $ are derived for this scheme with nonsmooth initial data. The convergence of the two schemes in the inhomogeneous case is also established. Finally, numerical experiments are performed to verify the theoretical results.

math.NA↗

arXiv2019

Asymptotic properties of the plane shear thickening fluids with bounded energy integral

In this note we investigate the asymptotic behavior of plane shear thickening fluids around a bounded obstacle. Different from the Navier-Stokes case considered by Gilbarg-Weinberger in \cite{GW1978}, where the good structure of the vorticity can be exploited and weighted energy estimates can be applied, we have to overcome the nonlinear term of high order. The decay estimates of the velocity was obtained by combining Point-wise Behavior Theorem in \cite{Galdi} and Brezis-Gallouet inequality in \cite{BG1980} together, which is independent of interest.

math.AP↗

arXiv2019

Learning a Layout Transfer Network for Context Aware Object Detection

We present a context aware object detection method based on a retrieve-and-transform scene layout model. Given an input image, our approach first retrieves a coarse scene layout from a codebook of typical layout templates. In order to handle large layout variations, we use a variant of the spatial transformer network to transform and refine the retrieved layout, resulting in a set of interpretable and semantically meaningful feature maps of object locations and scales. The above steps are implemented as a Layout Transfer Network which we integrate into Faster RCNN to allow for joint reasoning of object detection and scene layout estimation. Extensive experiments on three public datasets verified that our approach provides consistent performance improvements to the state-of-the-art object detection baselines on a variety of challenging tasks in the traffic surveillance and the autonomous driving domains.

cs.CV↗

arXiv2019

Merging External Bilingual Pairs into Neural Machine Translation

As neural machine translation (NMT) is not easily amenable to explicit correction of errors, incorporating pre-specified translations into NMT is widely regarded as a non-trivial challenge. In this paper, we propose and explore three methods to endow NMT with pre-specified bilingual pairs. Instead, for instance, of modifying the beam search algorithm during decoding or making complex modifications to the attention mechanism --- mainstream approaches to tackling this challenge ---, we experiment with the training data being appropriately pre-processed to add information about pre-specified translations. Extra embeddings are also used to distinguish pre-specified tokens from the other tokens. Extensive experimentation and analysis indicate that over 99% of the pre-specified phrases are successfully translated (given a 85% baseline) and that there is also a substantive improvement in translation quality with the methods explored here.

cs.CL↗

arXiv2019

Floquet-Induced Superfluidity with Periodically Modulated Interactions of Two-Species Hardcore Bosons in a One-dimensional Optical Lattice

We consider two species of hard-core bosons with density dependent hopping in a one-dimensional optical lattice, for which we propose experimental realizations using time-periodic driving. The quantum phase diagram for half-integer filling is determined by combining different advanced numerical simulations with analytic calculations. We find that a reduction of the density-dependent hopping induces a Mott-insulator to superfluid transition. For negative hopping a previously unknown state is found, where one species induces a gauge phase of the other species, which leads to a superfluid phase of gauge-paired particles. The corresponding experimental signatures are discussed.

cond-mat.str-el↗

arXiv2019

Quantum-to-classical correspondence in two-dimensional Heisenberg models

The quantum-to-classical correspondence (QCC) in spin models is a puzzling phenomenon where the static susceptibility of a quantum system agrees with its classical-system counterpart, at a different corresponding temperature, within the systematic error at a sub-percent level. We employ the bold diagrammatic Monte Carlo method to explore the universality of QCC by considering three different two-dimensional spin-1/2 Heisenberg models. In particular, we reveal the existence of QCC in two-parametric models.

cond-mat.str-el↗

arXiv2019

Weak measurement amplification based on thermal noise effect

Most studies for postselected weak measurement focus on using pure Gaussian state as a pointer, which can only give an amplification limit reaching the level of the ground state fluctuation. When the pointer is initialised in a thermal state, we find that the amplification limit after the postselection can reach the level of thermal fluctuation, indicating that the amplification effect achieving the level of thermal fluctuation is also increased with the temperature grow, and also give the amplification mechanism different from the one with pure Gaussian state pointer. To illustrate these results, we propose two schemes to implement room temperature amplification of the mechanical oscillator's displacement caused by a single photon in optomechanical system. The two schemes can both enhance the mechanical oscillator's original displacement by nearly seven orders of magnitude, attaining sensitivity to displacements of $ 0.26$ nm. Such amplification effect can be used to observe the impact of a single photon on a room temperature mechanical oscillator which is hard to detect in traditional measurement.

quant-ph↗

arXiv2019

Investigation of spin orbit torque driven dynamics in ferromagnetic heterostructures

We use time-resolved (TR) measurements based on the polar magneto-optical Kerr effect (MOKE) to study the magnetization dynamics excited by spin orbit torques in Py (Permalloy)/Pt and Ta/CoFeB bilayers. The analysis reveals that the field-like (FL) spin orbit torque (SOT) dominates the amplitude of the first oscillation cycle of the magnetization precession and the damping-like (DL) torque determines the final steady-state magnetization. In our bilayer samples, we have extracted the effective fields, hFL and hDL, of the two SOTs from the time-resolved magnetization oscillation spectrum. The extracted values are in good agreement with those extracted from time-integrated DCMOKE measurements, suggesting that the SOTs do not change at high frequencies. We also find that the amplitude ratio of the first oscillation to steady state is linearly proportional to the ratio hFL/hDL. The first oscillation amplitude is inversely proportional to, whereas the steady state value is independent of, the applied external field along the current direction.

cond-mat.mtrl-sci↗

arXiv2019

Scale MLPerf-0.6 models on Google TPU-v3 Pods

The recent submission of Google TPU-v3 Pods to the industry wide MLPerf v0.6 training benchmark demonstrates the scalability of a suite of industry relevant ML models. MLPerf defines a suite of models, datasets and rules to follow when benchmarking to ensure results are comparable across hardware, frameworks and companies. Using this suite of models, we discuss the optimizations and techniques including choice of optimizer, spatial partitioning and weight update sharding necessary to scale to 1024 TPU chips. Furthermore, we identify properties of models that make scaling them challenging, such as limited data parallelism and unscaled weights. These optimizations contribute to record performance in transformer, Resnet-50 and SSD in the Google MLPerf-0.6 submission.

cs.LG↗

arXiv2019

Color dependence of clustering of massive galaxies at 0.5$\le z \le$2.5: similar spatial distributions between green valley galaxies and AGNs

We present a measurement of the spatial clustering of rest-frame UV-selected massive galaxies at $0.5\le z \le 2.5$ in the COSMOS/UltraVISTA field. Considering four separate redshift bins with $Δz=0.5$, we construct three galaxy populations, i.e., red sequence (RS), blue cloud (BC), and green valley (GV) galaxies, according to their rest-frame extinction-corrected UV colors. The correlation lengths of these populations are confirmed to be dependent on their rest-frame UV color and redshift: UV redder galaxies are found to be more clustered. In all redshift bins, the GV galaxies generally have medium clustering amplitudes and are hosted within dark matter halos whose masses are more or less between those of RS and BC galaxies; and the clustering amplitude of GV galaxies is close to that of AGNs in the same redshift bin, suggesting that AGN activity may be responsible for transforming galaxy colors. After carefully examining their stellar masses, we find that the clustering amplitudes of galaxy samples with different colors are all similar once they have a similar median stellar mass and that the median stellar mass alone may be a good predictor of galaxy clustering.

astro-ph.GA↗