arXiv subjects

Lei Zhou

Publications and source records attributed to Lei Zhou.

At least 145 records · Page 8Linked to original sources

arXiv2019

ContextDesc: Local Descriptor Augmentation with Cross-Modality Context

Most existing studies on learning local features focus on the patch-based descriptions of individual keypoints, whereas neglecting the spatial relations established from their keypoint locations. In this paper, we go beyond the local detail representation by introducing context awareness to augment off-the-shelf local feature descriptors. Specifically, we propose a unified learning framework that leverages and aggregates the cross-modality contextual information, including (i) visual context from high-level image representation, and (ii) geometric context from 2D keypoint distribution. Moreover, we propose an effective N-pair loss that eschews the empirical hyper-parameter search and improves the convergence. The proposed augmentation scheme is lightweight compared with the raw local feature description, meanwhile improves remarkably on several large-scale benchmarks with diversified scenes, which demonstrates both strong practicality and generalization ability in geometric matching applications.

cs.CV↗

arXiv2019

Beyond Photometric Loss for Self-Supervised Ego-Motion Estimation

Accurate relative pose is one of the key components in visual odometry (VO) and simultaneous localization and mapping (SLAM). Recently, the self-supervised learning framework that jointly optimizes the relative pose and target image depth has attracted the attention of the community. Previous works rely on the photometric error generated from depths and poses between adjacent frames, which contains large systematic error under realistic scenes due to reflective surfaces and occlusions. In this paper, we bridge the gap between geometric loss and photometric loss by introducing the matching loss constrained by epipolar geometry in a self-supervised framework. Evaluated on the KITTI dataset, our method outperforms the state-of-the-art unsupervised ego-motion estimation methods by a large margin. The code and data are available at https://github.com/hlzz/DeepMatchVO.

cs.CV↗

arXiv2018

Matchable Image Retrieval by Learning from Surface Reconstruction

Convolutional Neural Networks (CNNs) have achieved superior performance on object image retrieval, while Bag-of-Words (BoW) models with handcrafted local features still dominate the retrieval of overlapping images in 3D reconstruction. In this paper, we narrow down this gap by presenting an efficient CNN-based method to retrieve images with overlaps, which we refer to as the matchable image retrieval problem. Different from previous methods that generates training data based on sparse reconstruction, we create a large-scale image database with rich 3D geometrics and exploit information from surface reconstruction to obtain fine-grained training data. We propose a batched triplet-based loss function combined with mesh re-projection to effectively learn the CNN representation. The proposed method significantly accelerates the image retrieval process in 3D reconstruction and outperforms the state-of-the-art CNN-based and BoW methods for matchable image retrieval. The code and data are available at https://github.com/hlzz/mirror.

cs.CV↗

arXiv2018

Nanoinstabilities of Cu2O porous nanostructured films as driven by nanocurvature effect and thermal activation effect

In this work, the instabilities at the nanoscale (i.e. nanoinstabilities) of triangular pyramids-like Cu2O porous nanostructured films (PNFs) are studied by heating treatments under different atmosphere and temperature. It is found that the nanoscale building triangular-pyramids turn round preferentially at the sharp angles and/or coalesce with their contacting ones by directional diffusion and plastic flow of atoms, which are driven by the nonuniformly-distributed surface nanocurvature. As a result, the triangular pyramids become quasi-sphere shape and the PNF evolves into a big, dense particles film. It is also observed that the heating or thermal activation effect efficiently promotes the reduction or oxidation of Cu2O pyramids and the crystallization or growth of the as-achieved Cu or CuO grains. The above physical and chemical instabilities or changes at the nanoscale of Cu2O PNFs can be well accounted for by the combined mechanism of nanocurvature effect and thermal activation effect. The nanocurvature effect can lower the energy barrier for the atom diffusion or plastic flow and lower the activation energy for the chemical reactions, while the thermal activation effect can supply the required kinetic energy or activation energy and make the atomic transportations and reactions kinetically possible. The findings reveal the evolution laws of morphology, crystal structure and composition of triangular pyramids-like Cu2O PNF during heating treatments, which can further be extended to other types of Cu2O PNFs. Also, the findings have important implications for the nanoinstabilities of Cu2O PNFs-based devices, especially those working at a high temperature.

cond-mat.mtrl-sci↗

arXiv2018

Orbital Stability of Earth Trojans

The only discovery of Earth Trojan 2010 TK$_7$ and the subsequent launch of OSIRIS-REx motive us to investigate the stability around the triangular Lagrange points $L_4$ and $L_5$ of the Earth. In this paper we present detailed dynamical maps on the $(a_0,i_0)$ plane with the spectral number (SN) indicating the stability. Two main stability regions, separated by a chaotic region arising from the $ν_3$ and $ν_4$ secular resonances, are found at low ($i_0\leq 15^\circ$) and moderate ($24^\circ\leq {i_0}\leq 37^\circ$) inclinations respectively. The most stable orbits reside below $i_0=10^\circ$ and they can survive the age of the Solar System. The nodal secular resonance $ν_{13}$ could vary the inclinations from $0^\circ$ to $\sim 10^\circ$ according to their initial values while $ν_{14}$ could pump up the inclinations to $\sim 20^\circ$ and upwards. The fine structures in the dynamical maps are related to higher-degree secular resonances, of which different types dominate different areas. The dynamical behaviour of the tadpole and horseshoe orbits, reflected in their secular precession, show great differences in the frequency space. The secular resonances involving the tadpole orbits are more sensitive to the frequency drift of the inner planets, thus the instabilities could sweep across the phase space, leading to the clearance of tadpole orbits. We are more likely to find terrestrial companions on horseshoe orbits. The Yarkovsky effect could destabilize Earth Trojans in varying degrees. We numerically obtain the formula describing the stabilities affected by the Yarkovsky effect and find the asymmetry between the prograde and retrograde rotating Earth Trojans. The existence of small primordial Earth Trojans that avoid being detected but survive the Yarkovsky effect for 4.5\,Gyr is substantially ruled out.

astro-ph.EP↗

arXiv2018

Simple property of heterogeneous aspiration dynamics: Beyond weak selection

How individuals adapt their behavior in cultural evolution remains elusive. Theoretical studies have shown that the update rules chosen to model individual decision making can dramatically modify the evolutionary outcome of the population as a whole. This hints at the complexities of considering the personality of individuals in a population, where each one uses its own rule. Here, we investigate whether and how heterogeneity in the rules of behavior update alters the evolutionary outcome. We assume that individuals update behaviors by aspiration-based self-evaluation and they do so in their own ways. Under weak selection, we analytically reveal a simple property that holds for any two-strategy multi-player games in well-mixed populations and on regular graphs: the evolutionary outcome in a population with heterogeneous update rules is the weighted average of the outcomes in the corresponding homogeneous populations, and the associated weights are the frequencies of each update rule in the heterogeneous population. Beyond weak selection, we show that this property holds for public goods games. Our finding implies that heterogeneous aspiration dynamics is additive. This additivity greatly reduces the complexity induced by the underlying individual heterogeneity. Our work thus provides an efficient method to calculate evolutionary outcomes under heterogeneous update rules.

physics.soc-ph↗

arXiv2018

Evolutionary multiplayer games on graphs with edge diversity

Evolutionary game dynamics in structured populations has been extensively explored in past decades. However, most previous studies assume that payoffs of individuals are fully determined by the strategic behaviors of interacting parties and social ties between them only serve as the indicator of the existence of interactions. This assumption neglects important information carried by inter-personal social ties such as genetic similarity, geographic proximity, and social closeness, which may crucially affect the outcome of interactions. To model these situations, we present a framework of evolutionary multiplayer games on graphs with edge diversity, where different types of edges describe diverse social ties. Strategic behaviors together with social ties determine the resulting payoffs of interactants. Under weak selection, we provide a general formula to predict the success of one behavior over the other. We apply this formula to various examples which cannot be dealt with using previous models, including the division of labor and relationship- or edge-dependent games. We find that labor division facilitates collective cooperation by decomposing a many-player game into several games of smaller sizes. The evolutionary process based on relationship-dependent games can be approximated by interactions under a transformed and unified game. Our work stresses the importance of social ties and provides effective methods to reduce the calculating complexity in analyzing the evolution of realistic systems.

q-bio.PE↗

arXiv2018

Aspiration dynamics generate robust predictions in structured populations

Evolutionary game dynamics in structured populations are strongly affected by updating rules. Previous studies usually focus on imitation-based rules, which rely on payoff information of social peers. Recent behavioral experiments suggest that whether individuals use such social information for strategy updating may be crucial to the outcomes of social interactions. This hints at the importance of considering updating rules without dependence on social peers' payoff information, which, however, is rarely investigated. Here, we study aspiration-based self-evaluation rules, with which individuals self-assess the performance of strategies by comparing own payoffs with an imaginary value they aspire, called the aspiration level. We explore the fate of strategies on population structures represented by graphs or networks. Under weak selection, we analytically derive the condition for strategy dominance, which is found to coincide with the classical condition of risk-dominance. This condition holds for all networks and all distributions of aspiration levels, and for individualized ways of self-evaluation. Our condition can be intuitively interpreted: one strategy prevails over the other if the strategy brings more satisfaction to individuals than the other does. Our work thus sheds light on the intrinsic difference between evolutionary dynamics induced by aspiration-based and imitation-based rules.

q-bio.PE↗

arXiv2018

Cu2O porous nanostructured films fabricated by positive bias sputtering deposition

In this work, the authors fabricated Cu2O porous nanostructured films (PNFs) on glass slide substrates by the newly developed positive bias deposition approach in a balanced magnetron sputtering (MS) system. It was found that the surface morphology, crystal structure and optical property of the as-deposited products were greatly dependent on the applied positive substrate bias. In particular, when the substrate was biased at +50 V and +150 V, both of the as-prepared Cu2O PNFs exhibited a unique triangular pyramids-like structure with obvious edges and corners and little gluing, a preferred orientation of (111) and a blue shift of energy band gap at 2.35 eV. Quantitative calculation results indicated that the traditional bombardment effects of electrons and sputtering argon ions were both negligible during the bias deposition in the balanced MS system. Instead, a new model of tip charging effect was further proposed to account for the controllable formation of PNFs by the balanced bias sputtering deposition.

cond-mat.mtrl-sci↗

arXiv2018

Observation of anomalous π modes in photonic Floquet engineering

Recent progresses on Floquet topological phases have shed new light on time-dependant quantum systems, among which one-dimensional (1D) Floquet systems have been under extensive theoretical research. However, an unambiguous experimental observation of these 1D Floquet topological phases has still been lacking. Here, by periodically bending ultrathin metallic arrays of coupled corrugated waveguides, a photonic Floquet simulator was well designed and successfully fabricated to simulate the periodically driven Su-Schrieffer-Heeger model. Intriguingly, under moderate driven frequencies, we first experimentally observed and theoretically verified the Floquet topological $π$ mode propagating along the array's boundary. The different evolutionary behaviors between static and non-static topological end modes have also been clearly demonstrated. Our experiment also reveals the universal high-frequency behavior in perically driven systems. We emphasize that, our system can serve as a powerful and versatile testing ground for various phenomena related to time-dependant 1D quantum phases, such as Thouless charge pumping and manybody localization.

quant-ph↗

arXiv2018

GeoDesc: Learning Local Descriptors by Integrating Geometry Constraints

Learned local descriptors based on Convolutional Neural Networks (CNNs) have achieved significant improvements on patch-based benchmarks, whereas not having demonstrated strong generalization ability on recent benchmarks of image-based 3D reconstruction. In this paper, we mitigate this limitation by proposing a novel local descriptor learning approach that integrates geometry constraints from multi-view reconstructions, which benefits the learning process in terms of data generation, data sampling and loss computation. We refer to the proposed descriptor as GeoDesc, and demonstrate its superior performance on various large-scale benchmarks, and in particular show its great success on challenging reconstruction tasks. Moreover, we provide guidelines towards practical integration of learned descriptors in Structure-from-Motion (SfM) pipelines, showing the good trade-off that GeoDesc delivers to 3D reconstruction tasks between accuracy and efficiency.

cs.CV↗

arXiv2018

Competition May Increase Social Happiness in Bipartite Matching Problem

Bipartite matching problem is to study two disjoint groups of agents who need to be matched pairwise. It can be applied to many real-world scenarios and explain many social phenomena. In this article, we study the effect of competition on bipartite matching problem by introducing correlated wish list. The results show that proper competition can improve the overall happiness of society and also reduce the instability of the matching result of unequal sized bipartite matching.

physics.soc-ph↗

arXiv2018

Fast Subspace Clustering Based on the Kronecker Product

Subspace clustering is a useful technique for many computer vision applications in which the intrinsic dimension of high-dimensional data is often smaller than the ambient dimension. Spectral clustering, as one of the main approaches to subspace clustering, often takes on a sparse representation or a low-rank representation to learn a block diagonal self-representation matrix for subspace generation. However, existing methods require solving a large scale convex optimization problem with a large set of data, with computational complexity reaches O(N^3) for N data points. Therefore, the efficiency and scalability of traditional spectral clustering methods can not be guaranteed for large scale datasets. In this paper, we propose a subspace clustering model based on the Kronecker product. Due to the property that the Kronecker product of a block diagonal matrix with any other matrix is still a block diagonal matrix, we can efficiently learn the representation matrix which is formed by the Kronecker product of k smaller matrices. By doing so, our model significantly reduces the computational complexity to O(kN^{3/k}). Furthermore, our model is general in nature, and can be adapted to different regularization based subspace clustering methods. Experimental results on two public datasets show that our model significantly improves the efficiency compared with several state-of-the-art methods. Moreover, we have conducted experiments on synthetic data to verify the scalability of our model for large scale datasets.

cs.LG↗

arXiv2018

Exploring Linear Relationship in Feature Map Subspace for ConvNets Compression

While the research on convolutional neural networks (CNNs) is progressing quickly, the real-world deployment of these models is often limited by computing resources and memory constraints. In this paper, we address this issue by proposing a novel filter pruning method to compress and accelerate CNNs. Our work is based on the linear relationship identified in different feature map subspaces via visualization of feature maps. Such linear relationship implies that the information in CNNs is redundant. Our method eliminates the redundancy in convolutional filters by applying subspace clustering to feature maps. In this way, most of the representative information in the network can be retained in each cluster. Therefore, our method provides an effective solution to filter pruning for which most existing methods directly remove filters based on simple heuristics. The proposed method is independent of the network structure, thus it can be adopted by any off-the-shelf deep learning libraries. Experiments on different networks and tasks show that our method outperforms existing techniques before fine-tuning, and achieves the state-of-the-art results after fine-tuning.

cs.CV↗

arXiv2018

A single-source molten salt synthesis of uniform octahedral Na2Ti3O7 particels composed of nanorods

Na2Ti3O7, as one of the most potential anode electrode materials, is expected to play an important role in various fields. In this work, we synthesized uniform hollow octahedral Na2Ti3O7 particles through directly sintering a kind of precursors obtained by (NH4)2TiF6 and NH4OH in the presence of sodium-containing cationic polyacrylamide. XRD, SEM, TEM were used to characterize precursors and Na2Ti3O7 particles. In addition, the synthesis conditions and the formation mechanism were discussed in detail. This single-source molten salt method is flexible, efficient and controllable. Moreover, the strategy of this method makes it possible for us to synthesize other sodium titanate materials.

cond-mat.mtrl-sci↗

arXiv2017

High-Speed Demodulation of weak FBGs Based on Microwave Photonics and Chromatic Dispersion

A high speed quasi-distributed demodulation method based on the microwave photonics and the chromatic dispersion effect is designed and implemented for weak fiber Bragg gratings (FBGs). Due to the effect of dispersion compensation fiber (DCF), FBG wavelength shift leads to the change of the difference frequency signal at the mixer. With the way of crossing microwave sweep cycle, all wavelengths of cascade FBGs can be high speed obtained by measuring the frequencies change. Moreover, through the introduction of Chirp-Z and Hanning window algorithm, the analysis of difference frequency signal is achieved very well. By adopting the single-peak filter as a reference, the length disturbance of DCF caused by temperature can be also eliminated. Therefore, the accuracy of this novel method is greatly improved, and high speed demodulation of FBGs can easily realize. The feasibility and performance are experimentally demonstrated using 105 FBGs with 0.1% reflectivity, 1 m spatial interval. Results show that each grating can be distinguished well, and the demodulation rate is as high as 40 kHz, the accuracy is about 8 pm.

physics.ins-det↗

arXiv2017

Face Parsing via a Fully-Convolutional Continuous CRF Neural Network

In this work, we address the face parsing task with a Fully-Convolutional continuous CRF Neural Network (FC-CNN) architecture. In contrast to previous face parsing methods that apply region-based subnetwork hundreds of times, our FC-CNN is fully convolutional with high segmentation accuracy. To achieve this goal, FC-CNN integrates three subnetworks, a unary network, a pairwise network and a continuous Conditional Random Field (C-CRF) network into a unified framework. The high-level semantic information and low-level details across different convolutional layers are captured by the convolutional and deconvolutional structures in the unary network. The semantic edge context is learnt by the pairwise network branch to construct pixel-wise affinity. Based on a differentiable superpixel pooling layer and a differentiable C-CRF layer, the unary network and pairwise network are combined via a novel continuous CRF network to achieve spatial consistency in both training and test procedure of a deep neural network. Comprehensive evaluations on LFW-PL and HELEN datasets demonstrate that FC-CNN achieves better performance over the other state-of-arts for accurate face labeling on challenging images.

cs.CV↗

arXiv2017

High-efficiency and full-space manipulation of electromagnetic wave-fronts with metasurfaces

Metasurfaces offered great opportunities to control electromagnetic (EM) waves, but currently available meta-devices typically work either in pure reflection or pure transmission mode, leaving half of EM space completely unexplored. Here, we propose a new type of metasurface, composed by specifically designed meta-atoms with polarization-dependent transmission and reflection properties, to efficiently manipulate EM waves in the full space. As a proof of concept, three microwave meta-devices are designed, fabricated and experimentally characterized. The first two can bend or focus EM waves at different sides (i.e., transmission/reflection sides) of the metasurfaces depending on the incident polarization, while the third one changes from a wave bender for reflected wave to a focusing lens for transmitted wave as the excitation polarization is rotated, with all these functionalities exhibiting very high efficiencies (in the range of 85%-91%). Our findings significantly expand the capabilities of metasurfaces in controlling EM waves, and can stimulate high-performance multi-functional meta-devices facing more challenging and diversified application demands.

physics.app-ph↗