arXiv subjects

Xin Fan

Publications and source records attributed to Xin Fan.

105 records · Page 6Linked to original sources

arXiv2018

Learning Aggregated Transmission Propagation Networks for Haze Removal and Beyond

Single image dehazing is an important low-level vision task with many applications. Early researches have investigated different kinds of visual priors to address this problem. However, they may fail when their assumptions are not valid on specific images. Recent deep networks also achieve relatively good performance in this task. But unfortunately, due to the disappreciation of rich physical rules in hazes, large amounts of data are required for their training. More importantly, they may still fail when there exist completely different haze distributions in testing images. By considering the collaborations of these two perspectives, this paper designs a novel residual architecture to aggregate both prior (i.e., domain knowledge) and data (i.e., haze distribution) information to propagate transmissions for scene radiance estimation. We further present a variational energy based perspective to investigate the intrinsic propagation behavior of our aggregated deep model. In this way, we actually bridge the gap between prior driven models and data driven networks and leverage advantages but avoid limitations of previous dehazing approaches. A lightweight learning framework is proposed to train our propagation network. Finally, by introducing a taskaware image separation formulation with a flexible optimization scheme, we extend the proposed model for more challenging vision tasks, such as underwater image enhancement and single image rain removal. Experiments on both synthetic and realworld images demonstrate the effectiveness and efficiency of the proposed framework.

cs.CV↗

arXiv2018

Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond

Blind image deblurring plays a very important role in many vision and multimedia applications. Most existing works tend to introduce complex priors to estimate the sharp image structures for blur kernel estimation. However, it has been verified that directly optimizing these models is challenging and easy to fall into degenerate solutions. Although several experience-based heuristic inference strategies, including trained networks and designed iterations, have been developed, it is still hard to obtain theoretically guaranteed accurate solutions. In this work, a collaborative learning framework is established to address the above issues. Specifically, we first design two modules, named Generator and Corrector, to extract the intrinsic image structures from the data-driven and knowledge-based perspectives, respectively. By introducing a collaborative methodology to cascade these modules, we can strictly prove the convergence of our image propagations to a deblurring-related optimal solution. As a nontrivial byproduct, we also apply the proposed method to address other related tasks, such as image interpolation and edge-preserved smoothing. Plenty of experiments demonstrate that our method can outperform the state-of-the-art approaches on both synthetic and real datasets.

cs.CV↗

arXiv2018

Toward Designing Convergent Deep Operator Splitting Methods for Task-specific Nonconvex Optimization

Operator splitting methods have been successfully used in computational sciences, statistics, learning and vision areas to reduce complex problems into a series of simpler subproblems. However, prevalent splitting schemes are mostly established only based on the mathematical properties of some general optimization models. So it is a laborious process and often requires many iterations of ideation and validation to obtain practical and task-specific optimal solutions, especially for nonconvex problems in real-world scenarios. To break through the above limits, we introduce a new algorithmic framework, called Learnable Bregman Splitting (LBS), to perform deep-architecture-based operator splitting for nonconvex optimization based on specific task model. Thanks to the data-dependent (i.e., learnable) nature, our LBS can not only speed up the convergence, but also avoid unwanted trivial solutions for real-world tasks. Though with inexact deep iterations, we can still establish the global convergence and estimate the asymptotic convergence rate of LBS only by enforcing some fairly loose assumptions. Extensive experiments on different applications (e.g., image completion and deblurring) verify our theoretical results and show the superiority of LBS against existing methods.

cs.CV↗

arXiv2017

Proximal Alternating Direction Network: A Globally Converged Deep Unrolling Framework

Deep learning models have gained great success in many real-world applications. However, most existing networks are typically designed in heuristic manners, thus lack of rigorous mathematical principles and derivations. Several recent studies build deep structures by unrolling a particular optimization model that involves task information. Unfortunately, due to the dynamic nature of network parameters, their resultant deep propagation networks do \emph{not} possess the nice convergence property as the original optimization scheme does. This paper provides a novel proximal unrolling framework to establish deep models by integrating experimentally verified network architectures and rich cues of the tasks. More importantly, we \emph{prove in theory} that 1) the propagation generated by our unrolled deep model globally converges to a critical-point of a given variational energy, and 2) the proposed framework is still able to learn priors from training data to generate a convergent propagation even when task information is only partially available. Indeed, these theoretical results are the best we can ask for, unless stronger assumptions are enforced. Extensive experiments on various real-world applications verify the theoretical convergence and demonstrate the effectiveness of designed deep models.

cs.CV↗

arXiv2017

Self-Reinforced Cascaded Regression for Face Alignment

Cascaded regression is prevailing in face alignment thanks to its accuracy and robustness, but typically demands manually annotated examples having low discrepancy between shape-indexed features and shape updates. In this paper, we propose a self-reinforced strategy that iteratively expands the quantity and improves the quality of training examples, thus upgrading the performance of cascaded regression itself. The reinforced term evaluates the example quality upon the consistence on both local appearance and global geometry of human faces, and constitutes the example evolution by the philosophy of "survival of the fittest". We train a set of discriminative classifiers, each associated with one landmark label, to prune those examples with inconsistent local appearance, and further validate the geometric relationship among groups of labeled landmarks against the common global geometry derived from a projective invariant. We embed this generic strategy into typical cascaded regressions, and the alignment results on several benchmark data sets demonstrate its effectiveness to predict good examples starting from a small subset.

cs.CV↗

arXiv2017

Sequential Dual Deep Learning with Shape and Texture Features for Sketch Recognition

Recognizing freehand sketches with high arbitrariness is greatly challenging. Most existing methods either ignore the geometric characteristics or treat sketches as handwritten characters with fixed structural ordering. Consequently, they can hardly yield high recognition performance even though sophisticated learning techniques are employed. In this paper, we propose a sequential deep learning strategy that combines both shape and texture features. A coded shape descriptor is exploited to characterize the geometry of sketch strokes with high flexibility, while the outputs of constitutional neural networks (CNN) are taken as the abstract texture feature. We develop dual deep networks with memorable gated recurrent units (GRUs), and sequentially feed these two types of features into the dual networks, respectively. These dual networks enable the feature fusion by another gated recurrent unit (GRU), and thus accurately recognize sketches invariant to stroke ordering. The experiments on the TU-Berlin data set show that our method outperforms the average of human and state-of-the-art algorithms even when significant shape and appearance variations occur.

cs.CV↗

arXiv2017

3D MOKE spin-orbit torque magnetometer

We demonstrate simultaneous detection of current driven antidamping-like and field-like spin-orbit torques in heavy metal/ferromagnetic metal bilayers by measuring all three magnetization components m_(x,) m_y, and m_z using the vector magneto-optic Kerr effect. We have also implemented a self-calibration method to accurately determine the effective fields of spin-orbit torques. With this technique, we investigate the magnitude and direction of spin-orbit torques in a series of platinum/permalloy samples. The values found are in excellent agreement with results obtained via quadratic magneto-optic Kerr effect, planar Hall effect, and spin transfer ferromagnetic resonance measurements.

physics.ins-det↗

arXiv2017

Mining Functional Modules by Multiview-NMF of Phenome-Genome Association

Background: Mining gene modules from genomic data is an important step to detect gene members of pathways or other relations such as protein-protein interactions. In this work, we explore the plausibility of detecting gene modules by factorizing gene-phenotype associations from a phenotype ontology rather than the conventionally used gene expression data. In particular, the hierarchical structure of ontology has not been sufficiently utilized in clustering genes while functionally related genes are consistently associated with phenotypes on the same path in the phenotype ontology. Results: We propose a hierarchal Nonnegative Matrix Factorization (NMF)-based method, called Consistent Multiple Nonnegative Matrix Factorization (CMNMF), to factorize genome-phenome association matrix at two levels of the hierarchical structure in phenotype ontology for mining gene functional modules. CMNMF constrains the gene clusters from the association matrices at two consecutive levels to be consistent since the genes are annotated with both the child phenotype and the parent phenotype in the consecutive levels. CMNMF also restricts the identified phenotype clusters to be densely connected in the phenotype ontology hierarchy. In the experiments on mining functionally related genes from mouse phenotype ontology and human phenotype ontology, CMNMF effectively improved clustering performance over the baseline methods. Gene ontology enrichment analysis was also conducted to reveal interesting gene modules. Conclusions: Utilizing the information in the hierarchical structure of phenotype ontology, CMNMF can identify functional gene modules with more biological significance than the conventional methods. CMNMF could also be a better tool for predicting members of gene pathways and protein-protein interactions. Availability: https://github.com/nkiip/CMNMF

cs.LG↗

arXiv2017

Observation of Spin-Orbit Effects with Spin Rotation Symmetry

The spin-orbit interaction enables interconversion between a charge current and a spin current. It is usually believed that in a nonmagnetic metal (NM) or at a NM/ferromagnetic metal (FM) bilayer interface, the symmetry of spin-orbit effects (SOE) requires that the spin current, charge current and spin orientation are all orthogonal to each other. Here we show the observation of a SOE near the NM/FM interface that exhibits a very different symmetry from the conventional spin Hall effect, insofar as the spin polarization is further rotated about the magnetization. These results imply that a perpendicularly polarized spin current can be generated with an in-plane charge current simply by use of a FM/NM bilayer with magnetization collinear to the charge current. The ability to generate a spin current with arbitrary polarization using typical magnetic materials will greatly benefit the development of magnetic memories.

cond-mat.mtrl-sci↗

arXiv2017

Spin-orbit torque switching of synthetic antiferromagnets

We report that synthetic antiferromagnets (SAFs) can be efficiently switched by spin-orbit torques (SOTs) and the switching scheme does not obey the usual SOT switching rule. We show that both the positive and negative spin Hall angle (SHA)-like switching can be observed in Pt/SAF structures with only positive SHA, depending on the strength of applied in-plane fields. A new switching mechanism directly arising from the asymmetric domain expansion is proposed to explain the anomalous switching behaviors. Contrary to the macrospin-based switching model that the SOT switching direction is determined by the sign of SHA, the new switching mechanism suggests that the SOT switching direction is dominated by the field-modulated domain wall motion and can be reversed even with the same sign of SHA. The new switching mechanism is further confirmed by the domain wall motion measurements. The anomalous switching behaviors provide important insights for understanding SOT switching mechanisms and also offer novel features for applications.

cond-mat.mes-hall↗

arXiv2016

A Fast Ellipse Detector Using Projective Invariant Pruning

Detecting elliptical objects from an image is a central task in robot navigation and industrial diagnosis where the detection time is always a critical issue. Existing methods are hardly applicable to these real-time scenarios of limited hardware resource due to the huge number of fragment candidates (edges or arcs) for fitting ellipse equations. In this paper, we present a fast algorithm detecting ellipses with high accuracy. The algorithm leverage a newly developed projective invariant to significantly prune the undesired candidates and to pick out elliptical ones. The invariant is able to reflect the intrinsic geometry of a planar curve, giving the value of -1 on any three collinear points and +1 for any six points on an ellipse. Thus, we apply the pruning and picking by simply comparing these binary values. Moreover, the calculation of the invariant only involves the determinant of a 3*3 matrix. Extensive experiments on three challenging data sets with 650 images demonstrate that our detector runs 20%-50% faster than the state-of-the-art algorithms with the comparable or higher precision.

cs.CV↗

arXiv2015

All-Optical Vector Measurement of Spin-Orbit-Induced Torques Using Both Polar and Quadratic Magneto-Optic Kerr Effects

We demonstrate that the magneto-optic-Kerr effect with normal light incidence can be used to obtain quantitative optical measurements of both components of spin-orbit-induced torque (both the antidamping and effective-field components) in heavy-metal/ferromagnet bilayers. This is achieved by analyzing the quadratic Kerr effect as well as the polar Kerr effect. The two effects can be distinguished by properly selecting the polarization of the incident light. We use this all-optical technique to determine the spin-orbit torques generated by a series of Pt/Permalloy samples, finding values in excellent agreement with spin-torque ferromagnetic resonance measurements.

cond-mat.mes-hall↗

arXiv2014

Attenuation correction with Region Growing Method used in the Positron Emission Mammography System

Positron Emission Mammography imaging system (PEMi) is a nuclear medicine diagnosis method dedicated for breast imaging. It provides a better resolution in detection of millimeter-sized breast tumors than whole-body PET. To address the requirement of semi-quantitative analysis with the radiotracer concentration map of the breast, a new attenuation correction method based on three-dimensional seeded region growing image segmentation (3DSRG-AC) solution was developed. The method gives a 3D connected region as the segmentation result instead of image slices. The continuously segmentation property makes this new method free of activity variation of breast tissues. Threshold value chosen is the key point for the segmentation process. The first valley of the grey level histogram of the reconstruction image is set as the lower threshold, which works fine in clinical application. Results show that attenuation correction for PEMi improves the image quality and the quantitative accuracy of radioactivity distribution determination. Attenuation correction also improves the probabilities to detect small and early breast tumors.

physics.med-ph↗

arXiv2014

A region segmentation based algorithm for building crystal position lookup table in scintillation detector

In scintillation detector, scintillation crystals are typically made into 2-dimension modular array. The location of incident gamma-ray need be calibrated due to spatial response nonlinearity. Generally, position histograms, the characteristic flood response of scintillation detectors, are used for position calibration. In this paper, a position calibration method based on crystal position lookup table which maps the inaccurate location calculated by Anger logic to the exact hitting crystal position has been proposed, Firstly, position histogram is segmented into disconnected regions. Then crystal marking points are labeled by finding the centroids of regions. Finally, crystal boundaries are determined and crystal position lookup table is generated. The scheme is evaluated by the whole-body PET scanner and breast dedicated SPECT detector developed by Institute of High Energy Physics, Chinese Academy of Sciences. The results demonstrate that the algorithm is accurate, efficient, robust and general purpose.

physics.ins-det↗

arXiv2013

A universal electromagnetic energy conversion adapter based on a metamaterial absorber

On the heels of metamaterial absorbers (MAs) which produce near perfect electromagnetic (EM) absorption and emission, we propose a universal electromagnetic energy conversion adapter (UEECA) based on MA. By choosing the appropriate energy converting sensors, the UEECA is able to achieve near 100% signal transfer ratio between EM energy and various forms of energy such as thermal, DC electric, or higher harmonic EM energy. The inherited subwavelength dimension and the EM field intensity enhancement can further empower UEECA in many critical applications such as energy harvesting, solar cell, and nonlinear optics. The principle of UEECA is understood with a transmission line model, which further provides a design strategy that can incorporate a variety of energy conversion devices. The concept is experimentally validated at a microwave frequency with a signal transfer ratio of 96% by choosing an RF diode as the energy converting sensor.

cond-mat.mtrl-sci↗