arXiv subjects

Chen Fang

Publications and source records attributed to Chen Fang.

At least 109 records · Page 6Linked to original sources

arXiv2019

Im2Pencil: Controllable Pencil Illustration from Photographs

We propose a high-quality photo-to-pencil translation method with fine-grained control over the drawing style. This is a challenging task due to multiple stroke types (e.g., outline and shading), structural complexity of pencil shading (e.g., hatching), and the lack of aligned training data pairs. To address these challenges, we develop a two-branch model that learns separate filters for generating sketchy outlines and tonal shading from a collection of pencil drawings. We create training data pairs by extracting clean outlines and tonal illustrations from original pencil drawings using image filtering techniques, and we manually label the drawing styles. In addition, our model creates different pencil styles (e.g., line sketchiness and shading style) in a user-controllable manner. Experimental results on different types of pencil drawings show that the proposed algorithm performs favorably against existing methods in terms of quality, diversity and user evaluations.

cs.CV↗

arXiv2019

Quasi 1D topological nodal vortex line phase in doped superconducting 3D Dirac Semimetals

We study the vortex bound states in three dimensional (3D) superconducting Dirac semimetals with time reversal symmetry. Assuming two Dirac points on the kz-axis and bulk s-wave superconductivity, with a quantum vortex line parallel to the z-direction, we find that the superconducting vortex line has a robust quasi-1D nodal phase. The nodal phase stems from the symmetry protected Dirac points in the normal state bands, and it can be characterized by a topological index (ν; n) at kz = 0 and kz = π, where νis the Z2 topological invariant for a 0D class-D system and n is the Z topological invariant for a 0D class-A system according to the Altland- Zirnbauer classification. Based on the topological index, we find that vortex end Majorana zero mode can coexist with the quasi-1D nodal phase in certain kinds of Dirac semimetals. The influence of the symmetry breaking perturbations on the quasi-1D nodal phase is also analyzed. Finally, we discuss the possible material realization of such nodal vortex line state.

cond-mat.supr-con↗

arXiv2019

Topological Vortex Phase Transitions in Iron-Based Superconductors

We study topological vortex phases in iron-based superconductors. Besides the previously known vortex end Majorana zero modes (MZMs) phase stemming from the existence of a three dimensional (3D) strong topological insulator state, we show that there is another topologically nontrivial phase as iron-based superconductors can be doped superconducting 3D weak topological insulators (WTIs). The vortex bound states in a superconducting 3D WTI exhibit two different types of quantum states, a robust nodal superconducting phase with pairs of bulk MZMs and a full-gap topologically nontrivial superconducting phase which has single vortex end MZM in a certain range of doping level. Moreover, we predict and summarize various topological phases in iron-based superconductors, and find that carrier doping and interlayer coupling can drive systems to have phase transitions between these different topological phases.

cond-mat.supr-con↗

arXiv2018

Topological states from topological crystals

We present a scheme to explicitly construct and classify general topological states jointly protected by an onsite symmetry group and a spatial symmetry group. We show that all these symmetry protected topological states can be adiabatically deformed (allowing for stacking of trivial states) into a special class of states we call topological crystals. A topological crystal in, for example, three dimensions is a real-space assembly of finite-sized pieces of topological states in one and two dimensions protected by the local symmetry group alone, arranged in a configuration invariant under the spatial group and glued together such there is no open edge or end. As a demonstration of principle, we explicitly enumerate all inequivalent topological crystals for non-interacting time-reversal symmetric electronic insulators with significant spin-orbit coupling and any one of the 230 space groups in three dimensions. Because every topological crystalline insulator can be deformed into a topological crystal, the enumeration of the latter gives topological crystalline insulators a full classification and for each class an explicit real-space construction. We also extend these results to give a unified classification including both strong topological insulators and topological crystalline insulators.

cond-mat.mes-hall↗

arXiv2018

Learning to Sketch with Deep Q Networks and Demonstrated Strokes

Doodling is a useful and common intelligent skill that people can learn and master. In this work, we propose a two-stage learning framework to teach a machine to doodle in a simulated painting environment via Stroke Demonstration and deep Q-learning (SDQ). The developed system, Doodle-SDQ, generates a sequence of pen actions to reproduce a reference drawing and mimics the behavior of human painters. In the first stage, it learns to draw simple strokes by imitating in supervised fashion from a set of strokeaction pairs collected from artist paintings. In the second stage, it is challenged to draw real and more complex doodles without ground truth actions; thus, it is trained with Qlearning. Our experiments confirm that (1) doodling can be learned without direct stepby- step action supervision and (2) pretraining with stroke demonstration via supervised learning is important to improve performance. We further show that Doodle-SDQ is effective at producing plausible drawings in different media types, including sketch and watercolor.

cs.CV↗

arXiv2018

Flow-Grounded Spatial-Temporal Video Prediction from Still Images

Existing video prediction methods mainly rely on observing multiple historical frames or focus on predicting the next one-frame. In this work, we study the problem of generating consecutive multiple future frames by observing one single still image only. We formulate the multi-frame prediction task as a multiple time step flow (multi-flow) prediction phase followed by a flow-to-frame synthesis phase. The multi-flow prediction is modeled in a variational probabilistic manner with spatial-temporal relationships learned through 3D convolutions. The flow-to-frame synthesis is modeled as a generative process in order to keep the predicted results lying closer to the manifold shape of real video sequence. Such a two-phase design prevents the model from directly looking at the high-dimensional pixel space of the frame sequence and is demonstrated to be more effective in predicting better and diverse results. Extensive experimental results on videos with different types of motion show that the proposed algorithm performs favorably against existing methods in terms of quality, diversity and human perceptual evaluation.

cs.CV↗

arXiv2018

All "Magic Angles" Are "Stable" Topological

We show that the electronic structure of the low-energy bands in the small angle-twisted bilayer graphene consists of a series of semi-metallic and topological phases. In particular we are able to prove, using an approximate low-energy particle-hole symmetry, that the gapped set of bands that exist around all magic angles has what we conjecture to be a stable topological index stabilized by a magnetic symmetry and reflected in the odd winding of the Wilson loop in the Moiré BZ. The approximate, emergent particle-hole symmetry is essential to the topology of graphene: when strongly broken, non-topological phases can appear. Our paper underpins topology as the crucial ingredient to the description of low-energy graphene. We provide a $4$-band short range tight-binding model whose $2$ lower bands have the same topology, symmetry, and flatness as those of the twisted graphene, and which can be used as an effective low-energy model. We then perform large-scale ($11000$ atoms per unit cell, 40 days per $\bf k$-point computing time) ab-initio calculations of a series of small angles, from $3^\circ$ to $1^\circ$, which show a more complex and somewhat qualitatively different evolution of the symmetry of the low-energy bands than that of the theoretical Moiré model, but which confirms the topological nature of the system. At certain angles, we find no insulating filling in graphene at $-4$ electrons per Moiré unit cell. The ab-initio evolution of gaps tends to differ from that of the continuum Moiré model.

cond-mat.mes-hall↗

arXiv2018

Quantitative mappings between symmetry and topology in solids

The study of spatial symmetries was accomplished during the last century, and had greatly improved our understanding of the properties of solids. Nowadays, the symmetry data of any crystal can be readily extracted from standard first-principles calculation. On the other hand, the topological data (topological invariants), the defining quantities of nontrivial topological states, are in general considerably difficult to obtain, and this difficulty has critically slowed down the search for topological materials. Here, we provide explicit and exhaustive mappings from symmetry data to topological data for arbitrary gapped band structure in the presence of time-reversal symmetry and any one of the 230 space groups. The mappings are completed using the theoretical tools of layer construction and symmetry-based indicators. With these results, finding topological invariants in any given gapped band structure reduces to a simple search in the mapping tables provided.

cond-mat.mes-hall↗

arXiv2018

Diagnosis for topological semimetals in the absence of spin-orbital coupling

Topological semimetals are under intensive theoretical and experimental studies. The first step of these studies is always the theoretical (numerical) predication of one of several candidate materials, starting from first principles. In these calculations, it is crucial that all topological band crossings, including their types and positions in the Brillouin zone, are found. While band crossings along high-symmetry lines, which are routinely scanned in numerics, are simple to locate, the ones at generic momenta are notoriously time-consuming to find, and may be easily missed. In this paper, we establish a theoretical scheme of diagnosis for topological semimetals where all band crossings are at generic momenta in systems with time-reversal symmetry and negligible spin-orbital coupling. The scheme only uses the symmetry (inversion and rotation) eigenvalues of the valence bands at high-symmetry points in the BZ as input, and provides the types, numbers and configurations of all topological band crossings, if any, at generic momenta. The nature of new diagnosis scheme allows for full automation and parallelizations, and paves way to high throughput numerical predictions of topological materials.

cond-mat.mes-hall↗

arXiv2018

"Factual" or "Emotional": Stylized Image Captioning with Adaptive Learning and Attention

Generating stylized captions for an image is an emerging topic in image captioning. Given an image as input, it requires the system to generate a caption that has a specific style (e.g., humorous, romantic, positive, and negative) while describing the image content semantically accurately. In this paper, we propose a novel stylized image captioning model that effectively takes both requirements into consideration. To this end, we first devise a new variant of LSTM, named style-factual LSTM, as the building block of our model. It uses two groups of matrices to capture the factual and stylized knowledge, respectively, and automatically learns the word-level weights of the two groups based on previous context. In addition, when we train the model to capture stylized elements, we propose an adaptive learning approach based on a reference factual model, it provides factual knowledge to the model as the model learns from stylized caption labels, and can adaptively compute how much information to supply at each time step. We evaluate our model on two stylized image captioning datasets, which contain humorous/romantic captions and positive/negative captions, respectively. Experiments shows that our proposed model outperforms the state-of-the-art approaches, without using extra ground truth supervision.

cs.CV↗

arXiv2018

Catalogue of Topological Electronic Materials

Topological electronic materials are new quantum states of matter hosting novel linear responses in the bulk and anomalous gapless states at the boundary, and are for scientific and applied reasons under intensive research in physics and in materials sciences. The detection for such materials has so far been hindered by the level of complication involved in the calculation of the so-called topological invariants, and is hence considered a specialized task that requires both experience with materials and expertise with advanced theoretical tools. Here we introduce an effective, efficient and fully automated algorithm in obtaining the topological invariants for all non-magnetic materials that are known to human, based on recently developed principles that allow for exhaustive mappings between the symmetry representation of occupied bands and the topological invariants. Our algorithm requires as input only the occupied-band information (energy and wavefunction) at a handful (up to eight) of high-symmetry points in the Brillouin zone, which is readily calculable with any first-principles software. In return, it is capable of providing a detailed topological classification of all non-magnetic materials. Equipped with this method we have scanned through a total of 39519 materials available in structural databases, and found that as many as 8056 of them are actually topological (8889 if spin-orbital coupling is neglected). These are further catalogued into classes of 5005 topological semimetals,1814 topological insulators and 1237 topological crystalline insulators, most of which are new to human knowledge. All the results are available and searchable at http://materiae.iphy.ac.cn/ .

cond-mat.mtrl-sci↗

arXiv2018

Dynamical Signature of Symmetry Fractionalization in Frustrated Magnets

The nontrivialness of quantum spin liquid (QSL) typically manifests in the non-local observables that signifies their existence, however, this fact actually casts shadow on detecting QSL with experimentally accessible probes. Here, we provide a solution by unbiasedly demonstrating dynamical signature of anyonic excitations and symmetry fractionalization in QSL. Employing large-scale quantum Monte Carlo simulation and stochastic analytic continuation, we investigate the extended XXZ model on the kagome lattice, and find out that across the phase transitions from Z2 QSLs to different symmetry breaking phases, spin spectral functions can reveal the presence and condensation of emergent anyonic spinon and vison excitations, in particular the translational symmetry fractionalization of the latter, which can be served as the unique dynamical signature of the seemingly ephemeral QSLs in spectroscopic techniques such as inelastic neutron or resonance (inelastic) X-ray scatterings.

cond-mat.str-el↗

arXiv2018

i-Josephson Junction as Topological Superconductor

We show that the time reversal symmetry inevitably breaks in a superconducting Josephson junction formed by two superconductors with different pairing symmetries dubbed as i-Josephson junction. While the leading conventional Josephson coupling vanishes in such an i-Josephson junction, the second order coupling from tunneling always generates chiral superconductivity orders with broken time reversal symmetry. Josephson frequency in the i-junction is doubled, namely $ω= 4eV /h$. The result provides a way to engineer topological superconductivity such as the d + id' -wave superconducting state characterized by a nonzero Chern number.

cond-mat.mes-hall↗

arXiv2018

Controllable growth of centimeter-size 2D perovskite heterostructural single crystals for highly narrow dual-band photodetectors

Two-dimensional (2D) organic-inorganic perovskites have recently attracted increasing attention due to their great environmental stability, remarkable quantum confinement effect and layered characteristic. Heterostructures consisting of 2D layered perovskites are expected to exhibit new physical phenomena inaccessible to the single 2D perovskites and can greatly extend their functionalities for novel electronic and optoelectronic applications. Herein, we develop a novel solution method to synthesize 2D perovskite single-crystals with the centimeter size, high phase purity, controllable junction depth, high crystalline quality and great stability for highly narrow dual-band photodetectors. On the basis of the different lattice constant, solubility and growth rate between different n number, the newly designed synthesis method allows to first grow n=1 perovskite guided by the self-assembled layer of the organic cations at the water-air interface and subsequently n=2 layer is formed via diffusion process. Such growth process provides an efficient away for us to readily obtain 2D perovskite heterostructural single-crystals with various thickness and junction depth by controlling the concentration, reaction temperature and time. Photodetectors based on such heterostructural single crystal plates exhibit extremely low dark current, high on-off current ratio, and highly narrow dual-band spectral response with a full-width at half-maximum of 20 nm at 540 nm and 34 nm at 610 nm. In particular, the synthetic strategy is general for other 2D perovskites and the narrow dual-band spectral response with all full-width at half-maximum below 40 nm can be continuously tuned from red to blue by properly changing the halide compositions.

physics.app-ph↗

arXiv2018

Symmetry Enforced Chiral Hinge States and Surface Quantum Anomalous Hall Effect in Magnetic Axion Insulator $\text{Bi}_{2-x}\text{Sm}_x\text{Se}_3$

A universal mechanism to generate chiral hinge states in the ferromagnetic axion insulator phase is proposed, which leads to an exotic transport phenomena, the quantum anomalous Hall effect (QAHE) on some particular surfaces determined by both the crystalline symmetry and the magnetization direction. A realistic material system Sm doped $\text{Bi}_2\text{Se}_3$ is then proposed to realize such exotic hinge states by combing the first principle calculations and the Green's function techniques. A physically accessible way to manipulate the surface QAHE is also proposed, which makes it very different from the QAHE in ordinary 2D systems.

cond-mat.mes-hall↗

arXiv2018

Speeding up Context-based Sentence Representation Learning with Non-autoregressive Convolutional Decoding

Context plays an important role in human language understanding, thus it may also be useful for machines learning vector representations of language. In this paper, we explore an asymmetric encoder-decoder structure for unsupervised context-based sentence representation learning. We carefully designed experiments to show that neither an autoregressive decoder nor an RNN decoder is required. After that, we designed a model which still keeps an RNN as the encoder, while using a non-autoregressive convolutional decoder. We further combine a suite of effective designs to significantly improve model efficiency while also achieving better performance. Our model is trained on two different large unlabelled corpora, and in both cases the transferability is evaluated on a set of downstream NLP tasks. We empirically show that our model is simple and fast while producing rich sentence representations that excel in downstream tasks.

cs.NE↗

arXiv2018

Visual to Sound: Generating Natural Sound for Videos in the Wild

As two of the five traditional human senses (sight, hearing, taste, smell, and touch), vision and sound are basic sources through which humans understand the world. Often correlated during natural events, these two modalities combine to jointly affect human perception. In this paper, we pose the task of generating sound given visual input. Such capabilities could help enable applications in virtual reality (generating sound for virtual scenes automatically) or provide additional accessibility to images or videos for people with visual impairments. As a first step in this direction, we apply learning-based methods to generate raw waveform samples given input video frames. We evaluate our models on a dataset of videos containing a variety of sounds (such as ambient sounds and sounds from people/animals). Our experiments show that the generated sounds are fairly realistic and have good temporal synchronization with the visual inputs.

cs.CV↗

arXiv2018

TextureGAN: Controlling Deep Image Synthesis with Texture Patches

In this paper, we investigate deep image synthesis guided by sketch, color, and texture. Previous image synthesis methods can be controlled by sketch and color strokes but we are the first to examine texture control. We allow a user to place a texture patch on a sketch at arbitrary locations and scales to control the desired output texture. Our generative network learns to synthesize objects consistent with these texture suggestions. To achieve this, we develop a local texture loss in addition to adversarial and content loss to train the generative network. We conduct experiments using sketches generated from real images and textures sampled from a separate texture database and results show that our proposed algorithm is able to generate plausible images that are faithful to user controls. Ablation studies show that our proposed pipeline can generate more realistic images than adapting existing methods directly.

cs.CV↗