Search

Nowadays, vision-based computing tasks play an important role in various real-world applications. However, many vision computing tasks, e.g. semantic segmentation, are usually computationally expensive, posing a challenge to the computing systems that are resource-constrained but require fast response speed. Therefore, it is valuable to develop accurate and real-time vision processing models that only require limited computational resources. To this end, we propose the Spatial-detail Guided Context Propagation Network (SGCPNet) for achieving real-time semantic segmentation. In SGCPNet, we propose the strategy of spatial-detail guided context propagation. It uses the spatial details of shallow layers to guide the propagation of the low-resolution global contexts, in which the lost spatial information can be effectively reconstructed. In this way, the need for maintaining high-resolution features along the network is freed, therefore largely improving the model efficiency. On the other hand, due to the effective reconstruction of spatial details, the segmentation accuracy can be still preserved. In the experiments, we validate the effectiveness and efficiency of the proposed SGCPNet model. On the Citysacpes dataset, for example, our SGCPNet achieves 69.5% mIoU segmentation accuracy, while its speed reaches 178.5 FPS on 768x1536 images on a GeForce GTX 1080 Ti GPU card. In addition, SGCPNet is very lightweight and only contains 0.61 M parameters., Comment: The paper has been accepted in the IEEE Transactions on Neural Networks and Learning Systems

Resistive random-access memory (ReRAM) is a promising candidate for the next generation non-volatile memory technology due to its simple read/write operations and high storage density. However, its crossbar array structure causes a severe interference effect known as the "sneak path." In this paper, we propose channel coding techniques that can mitigate both the sneak-path interference and the channel noise. The main challenge is that the sneak-path interference is data-dependent, and also correlated within a memory array, and hence the conventional error correction coding scheme will be inadequate. In this work, we propose an across-array coding strategy that assigns a codeword to multiple independent memory arrays, and exploit a real-time channel estimation scheme to estimate the instantaneous status of the ReRAM channel. Since the coded bits from different arrays experience independent channels, a "diversity" gain can be obtained during decoding, and when the codeword is adequately distributed over different memory arrays, the code actually performs as that over an uncorrelated channel. By performing decoding based on the scheme of treating-interference-as-noise (TIN), the ReRAM channel over different memory arrays is equivalent to a block varying channel we defined, for which we propose both the capacity bounds and a coding scheme. The proposed coding scheme consists of a serial concatenation of an optimized error correction code with a data shaper, which enables the ReRAM system to achieve a near capacity limit storage efficiency.

We investigate systematically the quark-hadron mixed phase in dense stellar matter, and its influence on compact star structures. The properties of quark matter and hadronic matter are fixed based on various model predictions. Beside adopting constant values, the surface tension $\Sigma$ for the quark-hadron interface is estimated with the multiple reflection expansion method and equivparticle model. To fix the structures of quark-hadron pasta phases, a continuous dimensionality of the structure is adopted as proposed by Ravenhall, Pethick, and Wilson. The corresponding properties of hybrid stars are then obtained and confronted with pulsar observations. It is found that the correlation between radius and tidal deformability in traditional neutron stars preserves in hybrid stars. For those permitted by pulsar observations, in almost all cases the quark phase persists inside the most massive compact stars. The quark-hadron interface plays an important role on hybrid star structures once quark matter emerges. The surface tension $\Sigma$ estimated with various methods increases with density, which predicts stiffer EOSs for the quark-hadron mixed phase and increases the maximum mass of hybrid stars. The EOSs of hybrid star matter are well constrained at densities $n\lesssim 0.8$ fm${}^{-3}$, while larger uncertainty is expected at higher densities.

Deformation field estimation is an important and challenging issue in many medical image registration applications. In recent years, deep learning technique has become a promising approach for simplifying registration problems, and has been gradually applied to medical image registration. However, most existing deep learning registrations do not consider the problem that when the receptive field cannot cover the corresponding features in the moving image and the fixed image, it cannot output accurate displacement values. In fact, due to the limitation of the receptive field, the 3 x 3 kernel has difficulty in covering the corresponding features at high/original resolution. Multi-resolution and multi-convolution techniques can improve but fail to avoid this problem. In this study, we constructed pyramidal feature sets on moving and fixed images and used the warped moving and fixed features to estimate their "residual" deformation field at each scale, called the Pyramidal Residual Deformation Field Estimation module (PRDFE-Module). The "total" deformation field at each scale was computed by upsampling and weighted summing all the "residual" deformation fields at all its previous scales, which can effectively and accurately transfer the deformation fields from low resolution to high resolution and is used for warping the moving features at each scale. Simulation and real brain data results show that our method improves the accuracy of the registration and the rationality of the deformation field.

Photocatalysis induced by sunlight is one of the most promising approach to environmental protection, solar energy conversion and sustainable production of fuels. The computational modeling of photocatalysis is a rapidly expending field which requires to adapt and further develop the available theoretical tools. The coupled transfer of proton and electron is an important reaction during photocatalysis. In this work, we present the first step of our methodology development in which we apply existing kinetic theory of such coupled transfer to a model system, namely, methanol photo-dissociation on rutile TiO$_2$(110) surface, with the help of high-level first-principles calculations. Moreover, we adapt the Stuchebrukhov-Hammes-Schiffer kinetic theory, where we use the Georgievskii-Stuchebrukhova vibronic coupling, to calculate the rate constant of the proton coupled electron transfer reaction for a particular pathway. In particular, we propose a modified expression to calculate the rate constant which enforces the near-resonance condition for the vibrational wavefunction during proton tunneling.

The symmetry energy obtained with the effective Skyrme energy density functional is related to the values of isoscalar effective mass and isovector effective mass, which is also indirectly related to the incompressibility of symmetric nuclear matter. In this work, we analyze the values of symmetry energy and its related nuclear matter parameters in five-dimensional parameter space by describing the heavy ion collision data, such as isospin diffusion data at 35 MeV/u and 50 MeV/u, neutron skin of $^{208}$Pb, and tidal deformability and maximum mass of neutron star. We obtain the parameter sets which can describe the isospin diffusion, neutron skin, tidal deformability and maximum mass of neutron star, and give the incompressibility $K_0$=250.23$\pm$20.16 MeV, symmetry energy coefficient $S_0$=31.35$\pm$2.08 MeV, the slope of symmetry energy $L$=59.57$\pm$10.06 MeV, isoscalar effective mass $m_s^*/m$=0.75$\pm$0.05 and quantity related to effective mass splitting $f_I$=0.005$\pm$0.170. At two times normal density, the symmetry energy we obtained is in 35-55 MeV. To reduce the large uncertainties of $f_I$, more critical works in heavy ion collisions at different beam energies are needed., Comment: 10 pages, 11 figures, 2 tables, accepted by Phys.Rev.C. This article draws heavily from arXiv:1911.05380. In the v1 version, we modified two typos in Table II

We investigate the stability and $e^+e^-$ pair creation of supercritically charged superheavy nuclei, $ud$QM nuggets, strangelets, and strangeon nuggets based on Thomas-Fermi approximation. The model parameters are fixed by reproducing their masses and charge properties reported in earlier publications. It is found that $ud$QM nuggets, strangelets, and strangeon nuggets may be more stable than ${}^{56}$Fe at $A\gtrsim 315$, $5\times10^4$, and $1.2\times10^8$, respectively. For those stable against neutron emission, the most massive superheavy element has a baryon number $\sim$965, while $ud$QM nuggets, strangelets, and strangeon nuggets need to have baryon numbers larger than $39$, 433, and $2.7\times10^5$. The $e^+e^-$ pair creation will inevitably start for superheavy nuclei with charge numbers $Z\geq177$, $ud$QM nuggets with $Z\geq163$, strangelets with $Z\geq 192$, and strangeon nuggets with $Z\geq 212$. A universal relation $Q/R_e = \left(m_e - \bar{\mu}_e\right)/\alpha$ is obtained at a given electron chemical potential $\bar{\mu}_e$, where $Q$ is the total charge and $R_e$ the radius of electron cloud. This predicts the maximum charge number by taking $\bar{\mu}_e=-m_e$. For supercritically charged objects with $\bar{\mu}_e<-m_e$, the decay rate for $e^+e^-$ pair production is estimated based on the JWKB approximation. It is found that most positrons are emitted at $t\lesssim 10^{-15}$ s, while a long lasting positron emission is observed for large objects with $R\gtrsim 1000$ fm. The emission and annihilation of positrons from supercritically charged objects may be partially responsible for the short $\gamma$-ray burst during the merger of binary compact stars, the 511 keV continuum emission, as well as the narrow faint emission lines in X-ray spectra from galaxies and galaxy clusters.

A superstructure can elicit versatile new properties of materials by breaking their original geometrical symmetries. It is an important topic in the layered graphene-like two-dimensional transition-metal dichalcogenides (TMDs), but its origin remains unclear. Using diamond-anvil cell techniques, synchrotron x-ray diffraction, x-ray absorption, and the first-principles calculations, we show that the evolution from the weak Van der Waals bonding to the Heisenberg covalent bonding between layers induces an isostructural transition in quasi-two-dimensional 1$\textit{T}$-type VSe$_2$ at high pressure. Furthermore, our results show that high-pressure induce a novel superstructure at 15.5 GPa, rather than suppress as it would normally, which is unexpected. It is driven by the Fermi surface nesting, enhanced by the pressure-induced distortion. The results suggest that the superstructure not only appears in the two-dimensional structure but also can emerge in the pressure-tuned three-dimensional structure with new symmetry and develop superconductivity., Comment: 17 pages, 5 figures

Choroid is the vascular layer of the eye, which is directly related to the incidence and severity of many ocular diseases. Optical Coherence Tomography (OCT) is capable of imaging both the cross-sectional view of retina and choroid, but the segmentation of the choroid region is challenging because of the fuzzy choroid-sclera interface (CSI). In this paper, we propose a biomarker infused global-to-local network (BioNet) for choroid segmentation, which segments the choroid with higher credibility and robustness. Firstly, our method trains a biomarker prediction network to learn the features of the biomarker. Then a global multi-layers segmentation module is applied to segment the OCT image into 12 layers. Finally, the global multi-layered result and the original OCT image are fed into a local choroid segmentation module to segment the choroid region with the biomarker infused as regularizer. We conducted comparison experiments with the state-of-the-art methods on a dataset (named AROD). The experimental results demonstrate the superiority of our method with $90.77\%$ Dice-index and 6.23 pixels Average-unsigned-surface-detection-error, etc., Comment: This paper has been cast for ISBI 2020

With the development of convolutional neural network, deep learning has shown its success for retinal disease detection from optical coherence tomography (OCT) images. However, deep learning often relies on large scale labelled data for training, which is oftentimes challenging especially for disease with low occurrence. Moreover, a deep learning system trained from data-set with one or a few diseases is unable to detect other unseen diseases, which limits the practical usage of the system in disease screening. To address the limitation, we propose a novel anomaly detection framework termed Sparsity-constrained Generative Adversarial Network (Sparse-GAN) for disease screening where only healthy data are available in the training set. The contributions of Sparse-GAN are two-folds: 1) The proposed Sparse-GAN predicts the anomalies in latent space rather than image-level; 2) Sparse-GAN is constrained by a novel Sparsity Regularization Net. Furthermore, in light of the role of lesions for disease screening, we present to leverage on an anomaly activation map to show the heatmap of lesions. We evaluate our proposed Sparse-GAN on a publicly available dataset, and the results show that the proposed method outperforms the state-of-the-art methods., Comment: Accepted to ISBI 2020

Retinal optical coherence tomography (OCT) and OCT angiography (OCTA) suffer from the degeneration of image quality due to speckle noise and bulk-motion noise, respectively. Because the cross-sectional retina has distinct features in OCT and OCTA B-scans, existing digital filters that can denoise OCT efficiently are unable to handle the bulk-motion noise in OCTA. In this Letter, we propose a universal digital filtering approach that is capable of minimizing both types of noise. Considering the retinal capillaries in OCTA are hard to differentiate in B-scans while having distinct curvilinear structures in 3D volumes, we decompose the volumetric OCT and OCTA data with 3D shearlets thus efficiently separate the retinal tissue and vessels from the noise in this transform domain. Compared with wavelets and curvelets, the shearlets provide better representation of the layer edges in OCT and the vasculature in OCTA. Qualitative and quantitative results show the proposed method outperforms the state-of-the-art OCT and OCTA denoising methods. Besides, the superiority of 3D denoising is demonstrated by comparing the 3D shearlet filtering with its 2D counterpart., Comment: This version has been accepted for publication in Opt. Lett

The detection of retinal vessel is of great importance in the diagnosis and treatment of many ocular diseases. Many methods have been proposed for vessel detection. However, most of the algorithms neglect the connectivity of the vessels, which plays an important role in the diagnosis. In this paper, we propose a novel method for retinal vessel detection. The proposed method includes a dense dilated network to get an initial detection of the vessels and a probability regularized walk algorithm to address the fracture issue in the initial detection. The dense dilated network integrates newly proposed dense dilated feature extraction blocks into an encoder-decoder structure to extract and accumulate features at different scales. A multiscale Dice loss function is adopted to train the network. To improve the connectivity of the segmented vessels, we also introduce a probability regularized walk algorithm to connect the broken vessels. The proposed method has been applied on three public data sets: DRIVE, STARE and CHASE_DB1. The results show that the proposed method outperforms the state-of-the-art methods in accuracy, sensitivity, specificity and also are under receiver operating characteristic curve.

Optical coherence tomography angiography (OCTA) requires high transverse sampling density for visualizing retinal and choroidal capillaries. Low transverse sampling causes resolution degradation, such as the angiograms in wide-field OCTA. In this paper, we propose to address this problem using deep learning. We conducted extensive experiments on converting the centrally cropped 3 x 3 mm2 field of view (FOV) of the 8 x 8 mm2 foveal OCTA images (a sampling density of 22.9 $\mu$m) to the native 3 x 3 mm2 en face OCTA images (a sampling density of 12.2 $\mu$m). We employed a cycle-consistent adversarial network architecture in this conversion. The quantitative analysis using the perceptual similarity measures shows the generated OCTA images are closer to the native 3 x 3 mm2 scans. Besides, the results show the proposed method could also enhance signal-to-noise ratio. We further applied our method to enhance diseased cases and calculate vascular biomarkers, which demonstrates its generalization performance and clinical perspective.