arXiv subjects

Hui Zeng

Publications and source records attributed to Hui Zeng.

70 records · Page 4Linked to original sources

arXiv2021

Real-Time Quantized Image Super-Resolution on Mobile NPUs, Mobile AI 2021 Challenge: Report

Image super-resolution is one of the most popular computer vision problems with many important applications to mobile devices. While many solutions have been proposed for this task, they are usually not optimized even for common smartphone AI hardware, not to mention more constrained smart TV platforms that are often supporting INT8 inference only. To address this problem, we introduce the first Mobile AI challenge, where the target is to develop an end-to-end deep learning-based image super-resolution solutions that can demonstrate a real-time performance on mobile or edge NPUs. For this, the participants were provided with the DIV2K dataset and trained quantized models to do an efficient 3X image upscaling. The runtime of all models was evaluated on the Synaptics VS680 Smart Home board with a dedicated NPU capable of accelerating quantized neural networks. The proposed solutions are fully compatible with all major mobile AI accelerators and are capable of reconstructing Full HD images under 40-60 ms while achieving high fidelity results. A detailed description of all models developed in the challenge is provided in this paper.

eess.IV↗

arXiv2020

Tuning the Magnetic and Electronic Properties of Strontium Titanate by Carbon Doping

The magnetic and electronic properties of strontium titanate with different carbon dopant configurations are explored using first-principles calculations with a generalized gradient approximation (GGA) and the GGA+U approach. Our results show that the structural stability, electronic properties and magnetic properties of C-doped SrTiO3 strongly depend on the distance between carbon dopants. In both GGA and GGA+U calculations, the doping structure is mostly stable with a nonmagnetic feature when the carbon dopants are nearest neighbors, which can be ascribed to the formation of a C-C dimer pair accompanied by stronger C-C and weaker C-Ti hybridizations as the C-C distance becomes smaller. As the C-C distance increases, C-doped SrTiO3 changes from an n-type nonmagnetic metal to ferromagnetic/antiferromagnetic half-metal and to an antiferromagnetic/ferromagnetic semiconductor in GGA calculations, while it changes from a nonmagnetic semiconductor to ferromagnetic half-metal and to an antiferromagnetic semiconductor using the GGA+U method. Our work demonstrates the possibility of tailoring the magnetic and electronic properties of C-doped SrTiO3, which might provide some guidance to extend the applications of strontium titanate as a magnetic or optoelectronic material.

cond-mat.mtrl-sci↗

arXiv2020

Learning Image-adaptive 3D Lookup Tables for High Performance Photo Enhancement in Real-time

Recent years have witnessed the increasing popularity of learning based methods to enhance the color and tone of photos. However, many existing photo enhancement methods either deliver unsatisfactory results or consume too much computational and memory resources, hindering their application to high-resolution images (usually with more than 12 megapixels) in practice. In this paper, we learn image-adaptive 3-dimensional lookup tables (3D LUTs) to achieve fast and robust photo enhancement. 3D LUTs are widely used for manipulating color and tone of photos, but they are usually manually tuned and fixed in camera imaging pipeline or photo editing tools. We, for the first time to our best knowledge, propose to learn 3D LUTs from annotated data using pairwise or unpaired learning. More importantly, our learned 3D LUT is image-adaptive for flexible photo enhancement. We learn multiple basis 3D LUTs and a small convolutional neural network (CNN) simultaneously in an end-to-end manner. The small CNN works on the down-sampled version of the input image to predict content-dependent weights to fuse the multiple basis 3D LUTs into an image-adaptive one, which is employed to transform the color and tone of source images efficiently. Our model contains less than 600K parameters and takes less than 2 ms to process an image of 4K resolution using one Titan RTX GPU. While being highly efficient, our model also outperforms the state-of-the-art photo enhancement methods by a large margin in terms of PSNR, SSIM and a color difference metric on two publically available benchmark datasets.

eess.IV↗

arXiv2020

The holographic p+ip solution failed to win the competition in dRGT massive gravity

In this paper, the holographic p-wave superfluid model with charged complex vector field is studied in dRGT massive gravity beyond the probe limit. The stability of p-wave and p+ip solutions are compared in the grand canonical ensemble. The p-wave solution always get lower value of grand potential than the p+ip solution, showing that the holographic system still favors an anisotropic (p-wave) solution even with considering a massive gravity theory in bulk. In the holographic superconductor models with dRGT massive gravity in bulk, a key sailing symmetry is found to be violated by fixing the reference metric parameter $c_0$. Therefore, in order to get the dependence of condensate and grand potential on temperature, different values of horizon radius should be considered in numerical work. With a special choice of model parameters, we further study the dependence of critical back-reaction strength on the graviton mass parameter, beyond which the superfluid phase transition become first order. We also give the dependence of critical temperature on the back reaction strength $b$ and graviton mass parameter $m^2$.

hep-th↗

arXiv2019

Grid Anchor based Image Cropping: A New Benchmark and An Efficient Model

Image cropping aims to improve the composition as well as aesthetic quality of an image by removing extraneous content from it. Most of the existing image cropping databases provide only one or several human-annotated bounding boxes as the groundtruths, which can hardly reflect the non-uniqueness and flexibility of image cropping in practice. The employed evaluation metrics such as intersection-over-union cannot reliably reflect the real performance of a cropping model, either. This work revisits the problem of image cropping, and presents a grid anchor based formulation by considering the special properties and requirements (e.g., local redundancy, content preservation, aspect ratio) of image cropping. Our formulation reduces the searching space of candidate crops from millions to no more than ninety. Consequently, a grid anchor based cropping benchmark is constructed, where all crops of each image are annotated and more reliable evaluation metrics are defined. To meet the practical demands of robust performance and high efficiency, we also design an effective and lightweight cropping model. By simultaneously considering the region of interest and region of discard, and leveraging multi-scale information, our model can robustly output visually pleasing crops for images of different scenes. With less than 2.5M parameters, our model runs at a speed of 200 FPS on one single GTX 1080Ti GPU and 12 FPS on one i7-6800K CPU. The code is available at: \url{https://github.com/HuiZeng/Grid-Anchor-based-Image-Cropping-Pytorch}.

cs.CV↗

arXiv2019

Reliable and Efficient Image Cropping: A Grid Anchor based Approach

Image cropping aims to improve the composition as well as aesthetic quality of an image by removing extraneous content from it. Existing image cropping databases provide only one or several human-annotated bounding boxes as the groundtruth, which cannot reflect the non-uniqueness and flexibility of image cropping in practice. The employed evaluation metrics such as intersection-over-union cannot reliably reflect the real performance of cropping models, either. This work revisits the problem of image cropping, and presents a grid anchor based formulation by considering the special properties and requirements (e.g., local redundancy, content preservation, aspect ratio) of image cropping. Our formulation reduces the searching space of candidate crops from millions to less than one hundred. Consequently, a grid anchor based cropping benchmark is constructed, where all crops of each image are annotated and more reliable evaluation metrics are defined. We also design an effective and lightweight network module, which simultaneously considers the region of interest and region of discard for more accurate image cropping. Our model can stably output visually pleasing crops for images of different scenes and run at a speed of 125 FPS. Code and dataset are available at: https://github.com/HuiZeng/Grid-Anchor-based-Image-Cropping.

cs.CV↗

arXiv2019

Toward Real-World Single Image Super-Resolution: A New Benchmark and A New Model

Most of the existing learning-based single image superresolution (SISR) methods are trained and evaluated on simulated datasets, where the low-resolution (LR) images are generated by applying a simple and uniform degradation (i.e., bicubic downsampling) to their high-resolution (HR) counterparts. However, the degradations in real-world LR images are far more complicated. As a consequence, the SISR models trained on simulated data become less effective when applied to practical scenarios. In this paper, we build a real-world super-resolution (RealSR) dataset where paired LR-HR images on the same scene are captured by adjusting the focal length of a digital camera. An image registration algorithm is developed to progressively align the image pairs at different resolutions. Considering that the degradation kernels are naturally non-uniform in our dataset, we present a Laplacian pyramid based kernel prediction network (LP-KPN), which efficiently learns per-pixel kernels to recover the HR image. Our extensive experiments demonstrate that SISR models trained on our RealSR dataset deliver better visual quality with sharper edges and finer textures on real-world scenes than those trained on simulated datasets. Though our RealSR dataset is built by using only two cameras (Canon 5D3 and Nikon D810), the trained model generalizes well to other camera devices such as Sony a7II and mobile phones.

cs.CV↗

arXiv2018

Liquid Epitaxial Growth of Two-dimensional Non-layer structured hybrid Perovskite

The hybrid perovskite CH3NH3PbX3 (X= Cl, Br, I) is a promising material for developing novel optoelectronic devices. Owing to the intrinsic non-layer structure, it remains challenging to synthesize molecularly thin CH3NH3PbX3 with large size. Here, we report a low-cost and highly efficient fabrication route to obtain large-scale single-crystalline 2D CH3NH3PbX3 perovskites on a mica substrate via liquid epitaxy. The 2D perovskite is characterized as 8 nm in thickness and hundreds of micrometers in lateral size. First-principles calculations suggest the strong potassium-halogen interactions at the perovskite/mica interface lower the interface energy of perovskites, driving their fast in-plane growth. Spectroscopic investigations reveal 2D CH3NH3PbBr3 possess small exciton binding energy of 30 meV, allowing a superior visible-light photodetector with a photoresponsivity of 126 A/W and a bandwidth exceeded 80 kHz. These features demonstrate that liquid epitaxy is a bottom-up approach to fabricate the non-layer structured 2D perovskites, which offer a new material platform for the device applications and fundamental investigations.

cond-mat.mtrl-sci↗

arXiv2018

Reduced bias nonparametric lifetime density and hazard estimation

Kernel-based nonparametric hazard rate estimation is considered with a special class of infinite-order kernels that achieves favorable bias and mean square error properties. A fully automatic and adaptive implementation of a density and hazard rate estimator is proposed for randomly right censored data. Careful selection of the bandwidth in the proposed estimators yields estimates that are more efficient in terms of overall mean squared error performance, and in some cases achieves a nearly parametric convergence rate. Additionally, rapidly converging bandwidth estimates are presented for use in second-order kernels to supplement such kernel-based methods in hazard rate estimation. Simulations illustrate the improved accuracy of the proposed estimator against other nonparametric estimators of the density and hazard function. A real data application is also presented on survival data from 13,166 breast carcinoma patients.

math.ST↗

arXiv2017

A Probabilistic Quality Representation Approach to Deep Blind Image Quality Prediction

Blind image quality assessment (BIQA) remains a very challenging problem due to the unavailability of a reference image. Deep learning based BIQA methods have been attracting increasing attention in recent years, yet it remains a difficult task to train a robust deep BIQA model because of the very limited number of training samples with human subjective scores. Most existing methods learn a regression network to minimize the prediction error of a scalar image quality score. However, such a scheme ignores the fact that an image will receive divergent subjective scores from different subjects, which cannot be adequately represented by a single scalar number. This is particularly true on complex, real-world distorted images. Moreover, images may broadly differ in their distributions of assigned subjective scores. Recognizing this, we propose a new representation of perceptual image quality, called probabilistic quality representation (PQR), to describe the image subjective score distribution, whereby a more robust loss function can be employed to train a deep BIQA model. The proposed PQR method is shown to not only speed up the convergence of deep model training, but to also greatly improve the achievable level of quality prediction accuracy relative to scalar quality score regression methods. The source code is available at https://github.com/HuiZeng/BIQA_Toolbox.

cs.CV↗

arXiv2017

Split degenerate states and stable p+ip phases from holography

In this paper, we investigate the p+$i$p superfluid phases in the complex vector field holographic p-wave model. We find that in the probe limit, the p+$i$p phase and the p-wave phase are equally stable, hence the p and $i$p orders can be mixed with an arbitrary ratio to form more general p+$λi$p phases, which are also equally stable with the p-wave and p+$i$p phases. As a result, the system possesses a degenerate thermal state in the superfluid region. We further study the case with considering the back reaction on the metric, and find that the degenerate ground states will be separated into p-wave and p+$i$p phases, and the p-wave phase is more stable. Finally, due to the different critical temperature of the zeroth order phase transitions from p-wave and p+$i$p phases to the normal phase, there is a temperature region where the p+$i$p phase exists but the p-wave phase doesn't. In this region we find the stable p+$i$p phase for the first time.

hep-th↗

arXiv2015

Phase transitions in a holographic s+p model with backreaction

In a previous paper (arXiv:1309.2204, JHEP 1311 (2013) 087), we present a holographic s+p superconductor model with a scalar triplet charged under an SU(2) gauge field in the bulk. We also study the competition and coexistence of the s-wave and p-wave orders in the probe limit. In this work we continue to study the model by considering the full back-reaction The model shows a rich phase structure and various condensate behaviors such as the "n-type" and "u-type" ones, which are also known as reentrant phase transitions in condensed matter physics. The phase transitions to the p-wave phase or s+p coexisting phase become first order in strong back-reaction cases. In these first order phase transitions, the free energy curve always forms a swallow tail shape, in which the unstable s+p solution can also play an important role. The phase diagrams of this model are given in terms of the dimension of the scalar order and the temperature in the cases of eight different values of the back reaction parameter, which show that the region for the s+p coexisting phase is enlarged with a small or medium back reaction parameter, but is reduced in the strong back-reaction cases.

hep-th↗

arXiv2015

P-T phase diagram of a holographic s+p model from Gauss-Bonnet gravity

In this paper, we study the holographic s+p model in 5-dimensional bulk gravity with the Gauss-Bonnet term. We work in the probe limit and give the $Δ$-T phase diagrams at three different values of the Gauss-Bonnet coefficient to show the effect of the Gauss-Bonnet term. We also construct the P-T phase diagrams for the holographic system using two different definitions of the pressure and compare the results.

hep-th↗

arXiv2013

Competition between the s-wave and p-wave superconductivity phases in a holographic model

We build a holographic superconductor model with a scalar triplet charged under an SU(2) gauge field in the bulk. In this model, the s-wave and p-wave condensates can be consistently realized. We find that there are totally four phases in this model, namely, the normal phase without any condensate, s-wave phase, p-wave phase and the s+p coexisting phase. By calculating Gibbs free energy, the s+p coexisting phase turns out to be thermodynamically favored once it can appear. The phase diagram with the dimension of the scalar operator and temperature is drawn. The temperature range for the s+p coexisting phase is very narrow, which shows the competition between the s-wave and p-wave orders in the superconductor model.

hep-th↗

arXiv2011

Looking for packing units of the protein structure

Lattice-model simulations and experiments of some small proteins suggest that folding is essentially controlled by a few conserved contacts. Residues of these conserved contacts form the minimum set of native contacts needed to ensure foldability. Keeping such conserved specific contacts in mind, we examine contacts made by two secondary structure elements of different helices or sheets and look for possible 'packing units' of the protein structure. Two short backbone fragments of width five centred at the C? atoms in contact is called an H-form, which serves as a candidate for the packing units. The structural alignment of protein family members or even across families indicates that there are conservative H-forms which are similar both in their sequences and local geometry, and consistent with the structural alignment. Carrying strong sequence signals, such packing units would provide 3D constraints as a complement of the potential functions for the structure prediction.

q-bio.BM↗

arXiv2009

The Interplay between Curvature Effect and Atomic Vacancy Defects in the Electronic Properties of Semi-Metallic Carbon Nanotubes

We investigate the electronic properties of semi-metallic (12,0) carbon nanotubes in the presence of a variety of mono-, di- and hexa-vacancy defects, by using first principle DFT combined with non-equilibrium Green's function technique. We show that defect states related to the vacancies hybridize with the extended states of the nanotubes to modify the band edge, and change the energy gap, resulting from the curvature effect. As a consequence, the nanotube conductance is not a monotonic function of the defect size and geometry. Paradoxically, hexa-vacancy nanotubes have higher conductance than di-vacancy nanotubes, which is due to the presence of mid-gap states originating from the defect, thereby enhancing the conductance.

cond-mat.mes-hall↗