Arbitrary Aspect Ratio Research Articles

Retargeting Video With an End-to-End Framework.

Video holds significance in computer graphics applications. Because of the heterogeneous of digital devices, retargeting videos becomes an essential function to enhance user viewing experience in such applications. In the research of video retargeting, preserving the relevant visual content in videos, avoiding flicking, and processing time are the vital challenges. Extending image retargeting techniques to the video domain is challenging due to the high running time. Prior work of video retargeting mainly utilizes time-consuming preprocessing to analyze frames. Plus, being tolerant of different video content, avoiding important objects from shrinking, and the ability to play with arbitrary ratios are the limitations that need to be resolved in these systems requiring investigation. In this paper, we present an end-to-end RETVI method to retarget videos to arbitrary aspect ratios. We eliminate the computational bottleneck in the conventional approaches by designing RETVI with two modules, content feature analyzer (CFA) and adaptive deforming estimator (ADE). The extensive experiments and evaluations show that our system outperforms previous work in quality and running time.

Equilibrium Flow Structure with Multiple Ion Species in Magnetically Confined Plasmas of an Arbitrary Aspect Ratio

Equilibrium Flow Structure with Multiple Ion Species in Magnetically Confined Plasmas of an Arbitrary Aspect Ratio

A semi-analytical wavelet finite element method for wave propagation in rectangular rods

Prior knowledge of the dispersion curves and mode shapes of guided waves provides valuable information for wave mode selection and excitation in the field of non-destructive evaluation (NDE) and structural health monitoring (SHM). They are typically computed by the matrix methods, the finite element (FE) and semi-analytical finite element (SAFE) methods. However, the former is prone to numerical instability, and the latter two are limited by the refinement level of the FE mesh. In this paper, a semi-analytical wavelet finite element (SAWFE) method is presented to characterize wave propagation in rectangular rods. The piecewise polynomial interpolation functions of the SAFE method are replaced by two-dimensional scaling functions of the B-spline wavelet on the interval (BSWI). To demonstrate the accuracy of the proposed SAWFE technique, the propagation of guided waves in an aluminium plate is studied first. Then, the propagation of guided waves in rectangular rods of arbitrary aspect ratio is investigated. The results of this work clearly show that the SAWFE method presented here has higher accuracy and efficiency than the SAFE method.

A data-driven model based on the numerical solution of the equivalent inclusion problem for the analysis of nonlinear short-fibre composites

The paper introduces a novel data-driven homogenization model developed for the computation of stress-strain curves for composites featuring arbitrary orientation distributions, aspect ratios, and volume fractions of fibres. The objective of the proposed model is to deliver both time-efficient and accurate results while taking into account a relatively small database. The database is generated through numerical solutions of the equivalent inclusion problem, encompassing results for predefined orientations and aspect ratios of the single inclusion (fibre). The database created for a specific matrix-fibre system can be employed by the surrogate model to predict outcomes for various cases involving different fibre configurations. The surrogate model comprises four key modules: aspect ratio interpolation, a discrete orientation averaging procedure involving a genetic algorithm, a spatial transformation procedure, and an iterative Mori-Tanaka scheme. The validity of the proposed data-driven model has been confirmed by comparing the obtained results with those acquired through analysis of the complex representative volume element (RVE) using Fast Fourier Transform (FFT)-based homogenization, as well as through comparisons with experimental results from the existing literature.

Resistive instabilities in general toroidal plasmas with neoclassical bootstrap currents

In this work, linear neoclassical resistive instabilities are investigated in general toroidal plasmas using standard perturbation theory. Using a neoclassical fluid model, we derive the singular layer equations modified by bootstrap currents and also obtain the dispersion relation of the resistive interchange mode and the neoclassical tearing mode (NTM), respectively. Additionally, we determine the stability criteria DRbs and Δcbs for bootstrap current-modified resistive modes. The resistive interchange mode is stable when DRbs<0 and the NTM is stable when Δ′<Δcbs, where Δ′ is the stability index of the tearing mode. It is found that, in tokamak plasmas with a positive magnetic shear, bootstrap currents have a destabilizing effect on resistive interchange modes, which not only increases the value of the stability criterion (DRbs) but also enhances the growth rate. In addition, bootstrap currents have a stabilizing effect on the growth rate of the NTM in a low growth rate region. However, bootstrap currents can also decrease the critical value Δcbs. In plasmas with negative magnetic shear, the opposite holds. Furthermore, the coupling effect between bootstrap currents and Pfirsch–Schlüter currents via magnetic field curvature is determined for the first time in this work. This coupling always has a stabilizing influence on the resistive interchange mode and can increase the value of Δcbs. The coupling is also independent of the sign of the magnetic shear and can be enhanced in low-aspect-ratio tokamaks (such as spherical tokamaks) or in plasma regions with low magnetic shear (as used in ITER hybrid scenarios). Our results are valid for low-n resistive instabilities in toroidal plasmas with arbitrary aspect ratios and β, where n is the toroidal mode number and β represents the ratio of the plasma pressure to the toroidal magnetic pressure. Overall, this investigation has broad parametric applications and deepens our understanding of the physical mechanisms underlying the influence of neoclassical effects on resistive instabilities.

Inertial migration of a neutrally buoyant spheroid in plane Poiseuille flow

We study the cross-stream inertial migration of a torque-free neutrally buoyant spheroid, of an arbitrary aspect ratio $\kappa$ , in wall-bounded plane Poiseuille flow for small particle Reynolds numbers ( $Re_p\ll 1$ ) and confinement ratios ( $\lambda \ll 1$ ), with the channel Reynolds number, $Re_c = Re_p/\lambda ^2$ , assumed to be arbitrary; here $\lambda =L/H$ , where $L$ is the semi-major axis of the spheroid and $H$ denotes the separation between the channel walls. In the Stokes limit ( $Re_p =0)$ , and for $\lambda \ll 1$ , a spheroid rotates along any of an infinite number of Jeffery orbits parameterized by an orbit constant $C$ , while translating with a time-dependent speed along a given ambient streamline. Weak inertial effects stabilize either the spinning ( $C=0$ ) or tumbling orbit ( $C=\infty$ ), or both, depending on $\kappa$ . The asymptotic separation of the Jeffery rotation and orbital drift time scales, from that associated with cross-stream migration, implies that migration occurs due to a Jeffery-averaged lift velocity. Although the magnitude of this averaged lift velocity depends on $\kappa$ and $C$ , the shape of the lift profiles are identical to those for a sphere, regardless of $Re_c$ . In particular, the equilibrium positions for a spheroid remain identical to the classical Segre–Silberberg ones for a sphere, starting off at a distance of about $0.6(H/2)$ from the channel centreline for small $Re_c$ , and migrating wallward with increasing $Re_c$ . For spheroids with $\kappa \sim O(1)$ , the Jeffery-averaged analysis is valid for $Re_p\ll 1$ ; for extreme aspect ratio spheroids, the regime of validity becomes more restrictive being given by $Re_p \kappa /\ln \kappa \ll 1$ and $Re_p/\kappa \ll 1$ for $\kappa \rightarrow \infty$ (slender fibres) and $\kappa \rightarrow 0$ (flat disks), respectively.

Bottom-up Filling of Nanosized Trenches with Silver to Fabricate Transparent Conducting Electrodes

Transparent conducting electrodes (TCEs) are essential in many optoelectronic devices including solar cells, sensors, displays, and LEDs. The use of transparent metal oxide layers, in particular Indium tin oxide (ITO), is the most common used approach. However, ITO comes with numerous drawbacks such as the relatively rarity of indium and hence high cost, fragility, strong UV absorption and relative high sheet resistance. An exciting alternative TCE is the use of interconnected metallic nanowires, where their sub-wavelength cross-sections makes them transparent to visible light, where especially silver nanowires are a good candidate because of their high conductivity.Here we show a promising cost-effective method called selective-area electrochemical deposition (SAEC) for the fabrication of these metal nanostructures. SAEC is based on a through-the-mask electrochemical deposition method, where the mask is made using substrate conformal imprint lithography. Next to the inherent low fabrication cost of the electrochemical deposition, it has numerous advantages such as bottom-up growth, scalability, operate at ambient conditions, control over nucleation, and yields high purity deposits. While submicron sized features have been already demonstrated by SAEC, obtaining homogeneous void free filling of features below 100 nm using direct plating is a major challenge because of the terminal effect and the poor control of nucleation of the seed layer.In this work, we have explored the SAEC fabrication of large area silver nanowire grids on ITO substrates and on TOPCon solar cells. We show that the nucleation density of the silver nanoparticles is the key parameter for the successful homogenous void-free filling of the trenches. The nucleation density namely determines the coalescence thickness, which sets a minimum thickness for the silver nanowires in order to be connected. We control the nucleation density of the silver nanoparticles by making use of the double pulse method, where a high overpotential nucleation pulse is applied for controlling the nucleation density. After nucleation, we merge the silver nuclei by applying a low overpotential growth pulse.After the successful filling of the trenches, we have compared the optical and electrical performance of the SAEC fabricated silver nanowire grids to those using conventional thermal or e-beam evaporation followed by lift-off. The SAEC-fabricated TCEs show high transmission (>96%) and extreme low resistance (as low as 3 Ω/sq), resulting into a superior figure of merit (FoM). Due to the bottom-up nature of this technique, arbitrary high aspect ratio nanowires can be achieved and therefore the relationship of decreasing transmission with decreasing sheet resistance is broken.Furthermore, we have also incorporated the silver NW grids in TOPCon cells and compared the performance to similar cells based on ITO. This work represents a low-cost and scalable strategy to fabricate sub 100 nm predefined metal structures, enabling the commercial viable fabrication of nanostructure based transparent conductive electrodes for optoelectronic devices.

Effective medium method in solution of the elastic and thermo-elastic homogenization problems for polycrystals with spheroidal transverse isotropic grains

Self-consistent effective medium method is used for determination of the effective elastic and thermo-elastic properties of polycrystalline materials with spheroidal transverse isotropic grains. The grains with arbitrary aspect ratios including very oblate (disks) and very prolate (fibers) spheroids are considered. The function of the Owens-March type is used for description of distribution of grains over orientations. The equations for the effective elastic stiffness tensor and the tensor of effective coefficients of thermo-expansion of polycrystals are derived using integral equations for elastic and thermo-elastic fields in heterogeneous media and the hypotheses of the effective medium method. The numerical algorithm of the method is described in detail. Numerical examples are presented for polycrystalline clay rock materials with strongly anisotropic grains and for polycrystalline zinc. Graphs of the dependences of the effective elastic constants and effective coefficients of thermo-expansion of the polycrystals on the parameters of the orientation distribution function and aspect ratios of grains are presented.

Near-resonance approximation of rotating Navier–Stokes equations

We formalise the concept of near resonance for the rotating Navier–Stokes equations, based on which we propose a novel way to approximate the original partial differential equation (PDE). The spatial domain is a three-dimensional flat torus of arbitrary aspect ratios. We prove that the family of proposed PDEs are globally well-posed for any rotation rate and initial datum of any size in any H s space with . Such approximations retain many more 3-mode interactions, and are thus more accurate, than the conventional exact-resonance approach. Our approach is free from any limiting argument that requires physical parameters to tend to zero or infinity, and is free from any use of small divisors (so that all estimates depend smoothly on the torus’s aspect ratios). The key estimate hinges on the counting of integer solutions of Diophantine inequalities rather than Diophantine equations. Using a range of novel ideas, we handle rigorously and optimally challenges arising from the non-trivial irrational functions in these inequalities. The main results and ingredients of the proofs can form part of the mathematical foundation of a non-asymptotic approach to nonlinear, oscillatory dynamics in real-world applications.

Pre-Activating Semantic Information for Image Aesthetic Assessment

Automatic image aesthetic evaluation is an attractive and challenging visual task. Recently, methods based on convolutional neural networks have achieved remarkable performance. However, semantic information, an intuitive prerequisite for evaluating image aesthetics, has not received enough attention regarding its importance in previous methods. How to efficiently extract semantic information and make better use of it to assist the aesthetic evaluation task remains unsolved. In this article, we propose to utilize the self-supervised model Auto-Encoder to extract semantic information in the form of multi-task learning. Then, a fusing module is prepended at the bottleneck layer to explicitly combine semantic information with aesthetic information in a pre-activated manner. Specifically, we implement a customized pooling operation to pool the semantic features extracted by Auto-Encoder and apply a weak constraint between the pooled semantic features and aesthetic information to realize the combination. The following regressor can complete aesthetic evaluation based on the semantic–aesthetic combined features. In addition, to enable our model to adapt to arbitrary aspect ratios of images, another pooling strategy called spatial pyramid pooling is adopted to obtain the image features of a fixed length. Our method achieves competitive performance on the public image aesthetic evaluation benchmark. Especially on the most commonly used metric Spearman rank-order correlation coefficient, the proposed model achieved the best performance compared with some state-of-the-art methods. Extensive ablation studies and visualization experiments were conducted to demonstrate the effectiveness of our method.

Photo Semantic Understanding and Retargeting by a Noise-Robust Regularized Topic Model

Retargeting aims at displaying a photo with an arbitrary aspect ratio, wherein the visually/semantically prominent objects are appropriately preserved and visual distortions can be well alleviated. Conventional retargeting models are built upon the visual perception of photos from a family of pre-specified communities ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g.</i> , “portrait”), wherein the underlying community-specific features are not learned explicitly. Thus they cannot appropriately retarget aerial photos, which contains a rich variety of objects with different scales. In this work, a novel aerial photo retargeting framework is designed by encoding the deep features from automatically detected Google Maps communities into a regularized probabilistic model. Specifically, we first propose an enhanced matrix factorization (MF) algorithm to calculate communities based on million-scale Google Maps pictures, for each of which deep feature is learned simultaneously. The enhanced MF incorporates label denoising, between-communities correlation, and deep feature encoding collaboratively. Subsequently, a probabilistic model called LTM is designed that quantifies the spatial layouts of multiple Google Maps communities in the underlying hidden space. To alleviate the overfitting from Google Maps communities with imbalanced numbers of aerial photos, a regularizer is added into the LTM. Finally, by leveraging the regularized LTM, we shrink the test photo horizontially/vertically to maximize the posterior probability of the retargted photo. Comprehensive subjective evaluations and visualizations have demonstrated the advantages of our method. Besides, our calculate Google Maps communities are competitively consistent with the ground truth, according to the quantitative comparisons on the 2M Google Maps photos.

Improved Analytical Triple-2D Leakage Inductance Model of Cone Winding Matrix Transformers

The transformer leakage inductance is one of the limiting factors for pulse shape quality in high voltage pulsed power (HVPP) converters that are essential in applications such as cancer treatment, particle accelerators, and free electron lasers. Cone winding matrix (CWM) transformers are commonly used in HVPP converters as they offer low leakage inductance and high insulation distance. This paper proposes an improved Triple-2D (iT2D) leakage inductance model for CWM transformers. The iT2D model identifies three basic cross sections of CWM transformers and calculates their associated leakage inductance contributions. The paper proposes an accurate and compact analytical formula of the leakage inductance contribution resulting from the magnetic energy stored in between the magnetic cores. For cross sections with cone windings, an expression is presented that computes the leakage inductance per unit length of non-parallel windings. The iT2D model is verified with measurements and finite element simulations of three transformer prototypes. The analytical solutions used in the iT2D model ensure applicability to arbitrary geometric aspect ratios and designs. Furthermore, the iT2D model is rapidly executable enabling its time-efficient integration in converter optimisations.

Energy Correlations of Non-Integrable Ising Models: The Scaling Limit in the Cylinder

We consider a class of non-integrable 2D Ising models whose Hamiltonian, in addition to the standard nearest neighbor couplings, includes additional weak multi-spin interactions which are even under spin flip. We study the model in cylindrical domains of arbitrary aspect ratio and compute the multipoint energy correlations at the critical temperature via a multiscale expansion, uniformly convergent in the domain size and in the lattice spacing. We prove that, in the scaling limit, the multipoint energy correlations converge to the same limiting correlations as those of the nearest neighbor Ising model in a finite cylinder with renormalized horizontal and vertical couplings, up to an overall multiplicative constant independent of the shape of the domain. The proof is based on a representation of the generating function of correlations in terms of a non-Gaussian Grassmann integral, and a constructive Renormalization Group (RG) analysis thereof. A key technical novelty compared with previous works is a systematic analysis of the effect of the boundary corrections to the RG flow, in particular a proof that the scaling dimension of boundary operators is better by one dimension than their bulk counterparts. In addition, a cancellation mechanism based on an approximate image rule for the fermionic Green’s function is of crucial importance for controlling the flow of the (superficially) marginal boundary terms under RG iterations.

Near resonant approximation of the rotating stratified Boussinesq system on a 3-torus

Based on a novel treatment of near resonances, we introduce a new approximation for the rotating stratified Boussinesq system on three-dimensional tori with arbitrary aspect ratios. The rotation and stratification parameters are arbitrary and not equal. We obtain global existence for the proposed nonlinear system for arbitrarily large initial data. This system is sufficiently accurate, with an important feature of coupling effects between slow and fast modes. The key to global existence is a sharp counting of the relevant number of nonlinear interactions. An additional regularity advantage arises from a careful examination of some mixed type interaction coefficients. In a wider context, the significance of our near resonant approach is a delicate balance between the inclusion of more interacting modes and the improvement of regularity properties, compared to the well-studied singular limit approach based on exact resonance.

Focusing on your subject: Deep subject-aware image composition recommendation networks

Photo composition is one of the most important factors in the aesthetics of photographs. As a popular application, composition recommendation for a photo focusing on a specific subject has been ignored by recent deep-learning-based composition recommendation approaches. In this paper, we propose a subject-aware image composition recommendation method, SAC-Net, which takes an RGB image and a binary subject window mask as input, and returns good compositions as crops containing the subject. Our model first determines candidate scores for all possible coarse cropping windows. The crops with high candidate scores are selected and further refined by regressing their corner points to generate the output recommended cropping windows. The final scores of the refined crops are predicted by a final score regression module. Unlike existing methods that need to preset several cropping windows, our network is able to automatically regress cropping windows with arbitrary aspect ratios and sizes. We propose novel stability losses for maximizing smoothness when changing cropping windows along with view changes. Experimental results show that our method outperforms state-of-the-art methods not only on the subject-aware image composition recommendation task, but also for general purpose composition recommendation. We also have designed a multistage labeling scheme so that a large amount of ranked pairs can be produced economically. We use this scheme to propose the first subject-aware composition dataset SACD, which contains 2777 images, and more than 5 million composition ranked pairs. The SACD dataset is publicly available at https://cg.cs.tsinghua.edu.cn/SACD/.

Bridge-over-water detection via modulated deformable convolution and attention mechanisms

Bridge-over-water detection plays vital role in urban surveillance and military reconnaissance. Bridges have arbitrary orientations and extreme aspect ratios in remote sensing images, and the preceding works cannot adequately extract bridge-related features. Small bridges are difficult to detect accurately in optical remote sensing images. The oriented bounding box annotations are required by previous deep-learning-based methods for detecting rotated objects. But obtaining the annotations is a laborious task. Though widely studied previously, they are still challenging problems. To address these problems, modulated deformable convolution and attention mechanisms were introduced in this paper. Modulated deformable convolution made the receptive field more flexible. The feature extraction capability of the network was enhanced. A new weighted structure was designed to quantify the contributions of channel and spatial attention mechanisms. A selective attention usage strategy was proposed to improve the detection performance. To locate bridge-over-water more precisely, a new bounding box conversion module was presented. There was no need for oriented bounding box annotations, and the process only relied on bridge-related prior knowledge. Multiple experiments were performed to verify the effectiveness of proposed methods.

Point RCNN: An Angle-Free Framework for Rotated Object Detection

Rotated object detection in aerial images is still challenging due to arbitrary orientations, large scale and aspect ratio variations, and extreme density of objects. Existing state-of-the-art rotated object detection methods mainly rely on angle-based detectors. However, angle-based detectors can easily suffer from a long-standing boundary problem. To tackle this problem, we propose a purely angle-free framework for rotated object detection, called Point RCNN. Point RCNN is a two-stage detector including both PointRPN and PointReg which are angle-free. Given an input aerial image, first, the backbone-FPN extracts hierarchical features, then, the PointRPN module generates an accurate rotated region of interests (RRoIs) by converting the learned representative points of each rotated object using the MinAreaRect function of OpenCV. Motivated by RepPoints, we designed a coarse-to-fine process to regress and refine the representative points for more accurate RRoIs. Next, based on the learned RRoIs of PointRPN, the PointReg module learns to regress and refine the corner points of each RRoI to perform more accurate rotated object detection. Finally, the final rotated bounding box of each rotated object can be attained based on the learned four corner points. In addition, aerial images are often severely unbalanced in categories, and existing rotated object detection methods almost ignore this problem. To tackle the severely unbalanced dataset problem, we propose a balanced dataset strategy. We experimentally verified that re-sampling the images of the rare categories can stabilize the training procedure and further improve the detection performance. Specifically, the performance was improved from 80.37 mAP to 80.71 mAP in DOTA-v1.0. Without unnecessary elaboration, our Point RCNN method achieved new state-of-the-art detection performance on multiple large-scale aerial image datasets, including DOTA-v1.0, DOTA-v1.5, HRSC2016, and UCAS-AOD. Specifically, in DOTA-v1.0, our Point RCNN achieved better detection performance of 80.71 mAP. In DOTA-v1.5, Point RCNN achieved 79.31 mAP, which significantly improved the performance by 2.86 mAP (from ReDet’s 76.45 to our 79.31). In HRSC2016 and UCAS-AOD, our Point RCNN achieved higher performance of 90.53 mAP and 90.04 mAP, respectively.

The rotation of a sedimenting spheroidal particle in a linearly stratified fluid

We derive analytically the angular velocity of a spheroid, of an arbitrary aspect ratio$\kappa$, sedimenting in a linearly stratified fluid. The analysis demarcates regions in parameter space corresponding to broadside-on and edgewise settling in the limit$Re, Ri_v \ll 1$, where$Re = \rho _0UL/\mu$and$Ri_v =\gamma L^3\,g/\mu U$, the Reynolds and viscous Richardson numbers, respectively, are dimensionless measures of the importance of inertial and buoyancy forces relative to viscous ones. Here,$L$is the spheroid semi-major axis,$U$an appropriate settling velocity scale,$\mu$the fluid viscosity and$\gamma \ (&gt;0)$the (constant) density gradient characterizing the stably stratified ambient, with the fluid density$\rho_0$taken to be a constant within the Boussinesq framework. A reciprocal theorem formulation identifies three contributions to the angular velocity: (1) an$O(Re)$inertial contribution that already exists in a homogeneous ambient, and orients the spheroid broadside-on; (2) an$O(Ri_v)$hydrostatic contribution due to the ambient stratification that also orients the spheroid broadside-on; and (3) a hydrodynamic contribution arising from the perturbation of the ambient stratification whose nature depends on$Pe$;$Pe = UL/D$being the Péclet number with$D$the diffusivity of the stratifying agent. For$Pe \ll 1$, this contribution is$O(Ri_v)$and orients prolate spheroids edgewise for all$\kappa \ (&gt;1)$. For oblate spheroids, it changes sign across a critical aspect ratio$\kappa _c \approx 0.41$, orienting oblate spheroids with$\kappa _c &lt; \kappa &lt; 1$edgewise and those with$\kappa &lt; \kappa _c$broadside-on. For$Pe \ll 1$, the hydrodynamic component is always smaller in magnitude than the hydrostatic one, so a sedimenting spheroid in this limit always orients broadside-on. For$Pe \gg 1$, the hydrodynamic contribution is dominant, being$O(Ri_v^{{2}/{3}}$) in the Stokes stratification regime characterized by$Re \ll Ri_v^{{1}/{3}}$, and orients the spheroid edgewise regardless of$\kappa$. Consideration of the inertial and large-$Pe$stratification-induced angular velocities leads to two critical curves which separate the broadside-on and edgewise settling regimes in the$Ri_v/Re^{{3}/{2}}$–$\kappa$plane, with the region between the curves corresponding to stable intermediate equilibrium orientations. The predictions for large$Pe$are broadly consistent with observations.

Improved Scaling Relationships for Seismic Moment and Average Slip of Strike-Slip Earthquakes Incorporating Fault-Slip Rate, Fault Width, and Stress Drop

ABSTRACT We develop a self-consistent scaling model relating magnitude Mw to surface rupture length (LE), surface displacement DE, and rupture width WE, for strike-slip faults. Knowledge of the long-term fault-slip rate SF improves magnitude estimates. Data are collected for 55 ground-rupturing strike-slip earthquakes that have geological estimates of LE, DE, and SF, and geophysical estimates of WE. We begin with the model of Anderson et al. (2017), which uses a closed form equation for the seismic moment of a surface-rupturing strike-slip fault of arbitrary aspect ratio and given stress drop, ΔτC. Using WE estimates does not improve Mw estimates. However, measurements of DE plus the relationship between ΔτC and surface slip provide an alternate approach to study WE. A grid of plausible stress drop and width pairs were used to predict displacement and earthquake magnitude. A likelihood function was computed from within the uncertainty ranges of the corresponding observed Mw and DE values. After maximizing likelihoods over earthquakes in length bins, we found the most likely values of WE for constant stress drop; these depend on the rupture length. The best-fitting model has the surprising form WE∝logLE—a gentle increase in width with rupture length. Residuals from this model are convincingly correlated to the fault-slip rate and also show a weak correlation with the crustal thickness. The resulting model thus supports a constant stress drop for ruptures of all lengths, consistent with teleseismic observation. The approach can be extended to test other observable factors that might improve the predictability of magnitude from a mapped fault for seismic hazard analyses.

Shuffling-SinGAN: Improvement on Generative Model from a Single Image

Recently, SinGAN takes the use of GANs into a new realm – unconditional generation learned from a single natural image. Following the SinGAN architecture, we propose Shuffling- SinGAN, an efficient unconditional generative model that trained on a single natural image for general image manipulation. Our new network includes a pyramid of fully convolutional GANs, in which each layer is responsible for learning the patch distribution at a different scale of the image. We can generate new samples with variability through our network, which have the function of maintaining the texture and global structure of the original image. New random image generated by the model after multiple training is different from the original image in detail. Inspired by sinIR, we decided to add random pixel shuffling to the network. After experimentation, we found that the changed model generated more random new samples. Shuffling-SinGAN allows generating new samples of arbitrary size and aspect ratio, that have significant variability, yet maintain both the global structure and the fine textures of the training image. User tests confirm that the generated samples are commonly confused to be real images. With quantitative evaluation, we show that Shuffling-SinGAN has competitive performance on random image generation.