Crisp Boundaries Research Articles

Learning Crisp Boundaries Using Deep Refinement Network and Adaptive Weighting Loss

Significant progress has been made in boundary detection with the help of convolutional neural networks. Recent boundary detection models not only focus on real object boundary detection but also “crisp” boundaries (precisely localized along the object's contour). There are two methods to evaluate crisp boundary performance. One uses more strict tolerance to measure the distance between the ground truth and the detected contour. The other focuses on evaluating the contour map without any postprocessing. In this study, we analyze both methods and conclude that both methods are two aspects of crisp contour evaluation. Accordingly, we propose a novel network named deep refinement network (DRNet) that stacks multiple refinement modules to achieve richer feature representation and a novel loss function, which combines cross-entropy and dice loss through effective adaptive fusion. Experimental results demonstrated that we achieve state-of-the-art performance for several available datasets.

Nonlinear homogenization for topology optimization

Non-linear homogenization of hyperelastic materials is reviewed and adapted to topology optimization. The homogenization is based on the method of multiscale virtual power in which the unit cell is subjected to either macroscopic deformation gradients or equivalently to Bloch type displacement boundary conditions. A detailed discussion regarding domain symmetry of the unit cell and its effect on uniaxial loading conditions is provided. The density approach is used to formulate the topology optimization problem which is solved via the method of moving asymptotes. The adjoint sensitivity analysis considers response functions that quantify both the displacement and incremental displacement. Notably, the transfer of the sensitivities from the microscale to the macroscale is presented in detail. A periodic filter and thresholding are used to regularize the topology optimization problem and to generate crisp boundaries. The proposed methodology is used to design hyperelastic microstructures comprised of Neo-Hookean constituents for maximum load carrying capacity subject to negative Poisson’s ratio constraints.

MATNet: Motion-Attentive Transition Network for Zero-Shot Video Object Segmentation.

In this paper, we present a novel end-to-end learning neural network, i.e., MATNet, for zero-shot video object segmentation (ZVOS). Motivated by the human visual attention behavior, MATNet leverages motion cues as a bottom-up signal to guide the perception of object appearance. To achieve this, an asymmetric attention block, named Motion-Attentive Transition (MAT), is proposed within a two-stream encoder network to firstly identify moving regions and then attend appearance learning to capture the full extent of objects. Putting MATs in different convolutional layers, our encoder becomes deeply interleaved, allowing for close hierarchical interactions between object apperance and motion. Such a biologically-inspired design is proven to be superb to conventional two-stream structures, which treat motion and appearance independently in separate streams and often suffer severe overfitting to object appearance. Moreover, we introduce a bridge network to modulate multi-scale spatiotemporal features into more compact, discriminative and scale-sensitive representations, which are subsequently fed into a boundary-aware decoder network to produce accurate segmentation with crisp boundaries. We perform extensive quantitative and qualitative experiments on four challenging public benchmarks, i.e., DAVIS16, DAVIS17, FBMS and YouTube-Objects. Results show that our method achieves compelling performance against current state-of-the-art ZVOS methods. To further demonstrate the generalization ability of our spatiotemporal learning framework, we extend MATNet to another relevant task: dynamic visual attention prediction (DVAP). The experiments on two popular datasets (i.e., Hollywood-2 and UCF-Sports) further verify the superiority of our model. Our implementations have been made publicly available at https://github.com/tfzhou/MATNet.

Weakly supervised deep network for spatiotemporal localization and detection of human actions in wild conditions

Human action localization in any long, untrimmed video can be determined from where and what action takes place in a given video segment. The main hurdles in human action localization are the spatiotemporal randomnesses of their happening in a parallel mode which means the location (a particular set of frames containing action instances) and duration of any particular action in real-life video sequences generally are not fixed. At another end, the uncontrolled conditions such as occlusions, viewpoints and motions at the crisp boundary of the action sequences demand to develop a fast deep network which can be easily trained from unlabeled samples of complex video sequences. Motivated from the facts, we proposed a weakly supervised deep network model for human action localization. The model is trained from unlabeled action samples from UCF50 action benchmark. The five-channel data obtained from the concatenation of RGB (three-channel) and optical flow vectors (two-channel) are fed to the proposed convolutional neural network. LSTM network is used to yield the region of action happening area. The performance of the model is tested on UCF-sports dataset. The observation and comparative results reflect that our model can localize any action from annotation-free data samples captured in uncontrolled conditions.

Disaggregation of conventional soil map by generating multi realizations of soil class distribution (case study: Saadat Shahr plain, Iran).

Conventional soil maps generally depict information about soil spatial distribution in the framework of crisp boundaries of tessellated soil polygons. Such maps in standard soil survey procedures determine map unit composition on the basis of relative acreage occupied by individual major and minor soil components within soil map unit without addressing the specific location of each component in the polygon boundary. These limitations in addition to the sharp-edge boundaries of conventional soil maps are considered obstacles for modern land resource management. To increase detail in the polygon of conventional soil maps, we have produced a spatially disaggregated soil class map of a relatively flat agricultural plain called Saadat Shahr in South-Central Iran, using DSMART algorithm. DSMART is a known DSM-based disaggregation and harmonization algorithm that works through resampled classification trees to estimate the probability of the existence of each possible soil classes and also to prepare the maps of the most probable soil class, second most probable, and so on in raster format. The conventional soil map and 124 georeferenced profiles, as well as a set of numerical and categorical auxiliary data in 10-m resolutions in the extent of the study area utilized as the SCORPAN variables, were used as the inputs of the DSMART algorithm. A set of 78 independent sampling points generated by Latin hypercube sampling scheme were investigated and then used for validation of the DSMART raster outputs. The results indicated an improvement in disaggregated maps in the case of allocating soil components within the map units. In the generated DSMART, overall accuracy for seven soil subgroups was 68%. The best prediction obtained for Typic Xerorthents and Typic Calcixerepts, meanwhile a few classes were poorly predicted. For second most probable and third most probable maps, 17% and 0.5% of predicted soils match that observed respectively. This study revealed that DSMART as a disaggregation method can be used for enhancing existence soil map with poor descriptive data in the case of allocating soil classes in a more detailed way compared to the relevant original map.

Recent glacier and lake changes in High Mountain Asia and their relation to precipitation changes

Abstract. We present an updated, spatially resolved estimate of 2003–2008 glacier surface elevation changes for the entire region of High Mountain Asia (HMA) from ICESat laser altimetry data. The results reveal a diverse pattern that is caused by spatially greatly varying glacier sensitivity, in particular to precipitation availability and changes. We introduce a spatially resolved zonation where ICESat samples are grouped into units of similar glacier behaviour, glacier type and topographic settings. In several regions, our new zonation reveals local differences and anomalies that have not been described previously. Glaciers in the Eastern Pamirs, Kunlun Shan and central TP were thickening by 0.1–0.7 m a−1, and the thickening anomaly has a crisp boundary in the Eastern Pamirs that continues just north of the central Karakoram. Glaciers in the south and east of the TP were thinning, with increasing rates towards southeast. We attribute the glacier thickening signal to a stepwise increase in precipitation around ∼1997–2000 on the Tibetan Plateau (TP). The precipitation change is reflected by growth of endorheic lakes in particular in the northern and eastern TP. We estimate lake volume changes through a combination of repeat lake extents from Landsat data and shoreline elevations from ICESat and the Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) for over 1300 lakes. The rise in water volume contained in the lakes corresponds to 4–25 mm a−1, when distributed over entire catchments, for the areas where we see glacier thickening. The precipitation increase is also visible in sparse in situ measurements and MERRA-2 climate reanalysis data but less visible in ERA-Interim reanalysis data. Taking into account evaporation loss, the difference between average annual precipitation during the 1990s and 2000s suggested by these datasets is 34–100 mm a−1, depending on region, which can fully explain both lake growth and glacier thickening (Kunlun Shan) or glacier geometry changes such as thinning tongues while upper glacier areas were thickening or stable (eastern TP). The precipitation increase reflected in these glacier changes possibly extended to the northern slopes of the Tarim Basin, where glaciers were nearly in balance in 2003–2008. Along the entire Himalaya, glaciers on the first orographic ridge, which are exposed to abundant precipitation, were thinning less than glaciers in the dryer climate of the inner ranges. Thinning rates in the Tien Shan vary spatially but are rather stronger than in other parts of HMA.

Fuzzy-Based Intelligent Algorithm for Diagnosis of Drive Faults in Induction Motor Drive System

This paper presents the design and development of a novel, fuzzy-based algorithm for the detection and diagnosis of drive faults in an induction motor drive system (IMDS). A detailed investigation on the performance of IMDS under various faults and load conditions revealed that the combination of root-mean-square value and total harmonic distortion (THD) of the stator currents can accurately transpire various fault conditions. In this work, the efficacy of fuzzy logic is employed to characterize and diagnose the fault since it is difficult to find crisp boundaries for the correlation between the extracted parameters and fault conditions. The performance of the developed algorithm is tested and verified using simulation in MATLAB Simulink.

Gerrymandering & Justiciability: The Political Question Doctrine After Gill v. Whitford

In Rucho v. Common Cause, the Supreme Court held that partisan-gerrymandering claims present political questions beyond the jurisdiction of the federal courts. The decision marks the first time that a majority of the Court has used the political question doctrine to dismiss a claim based on the supposed unmanageability of legal standards alone. Despite this development being essential to the Court’s holding, its significance went largely unexplored and uncontested. This Article deconstructs Rucho’s articulation and application of the political question doctrine and makes two contributions. First, the Article disentangles the political question doctrine from neighboring justiciability doctrines. The result is a set of substantive principles that should guide federal courts as they exercise a range of routine judicial functions—remedial, adjudicative, and interpretive. Rather than unrealistically attempting to draw crisp jurisdictional boundaries between exercises of “political” and “judicial” power, the political question doctrine should seek to moderate their inevitable (and frequent) clash. Standing doctrine should continue to guide courts in determining whether they have authority over a case involving a political question. But the political question doctrine should guide courts in determining how to navigate areas of overlapping constitutional authority covered by the question. Second, the Article challenges the assumption that manageability can and should provide a standalone test of jurisdiction. Rucho’s “manageability exception” to Article III actively undermines the separation of powers principles it purports to protect. Treating manageability as an independent test of jurisdiction invites federal courts to opine about abstract questions not presented by the case or controversy before them and to abdicate their duty to decide cases properly before them without any constitutional command to the contrary. Rucho offers a profound illustration. Chief Justice Roberts asserts that partisan-gerrymandering claims are not “resolvable according to legal principles,” but his opinion fails to set out any legal standard rooted in the Constitution for deciding when a claim is or is not “manageable.” The result is a decision as hollow as it is hypocritical. This Article proposes a new approach to the political question doctrine, explains where the Court went wrong in Rucho, and explores what it could mean for the future of partisan-gerrymandering claims and justiciability doctrine alike.

Granular Mining and Big Data Analytics: Rough Models and Challenges

Data analytics in granular computing framework is considered for several mining applications, such as in video analysis, bioinformatics and online social networks which have all the characteristics of Big data. The role of granulation, lower approximation and r–f information measure is exhibited. While the lower approximation over a video sequence signifies the object model for unsupervised tracking, it characterizes the probability (relative frequency) of definite regions in ranking miRNAs for normal and cancer classification. For neural learning, the information on definite region is used as the initial knowledge for encoding while generating the networks through evolution. Granules considered are of different sizes and dimensions with fuzzy and crisp boundaries. The tracking method is effective in handling different ambiguous situations, e.g., overlapping objects, newly appeared object(s), multiple objects in different directions and speeds, in unsupervised mode. The ranking algorithm could find only 1% miRNAs to result in significantly higher F-score than the entire set. Fuzzy–rough communities detected over the granular model of social networks are suitable in dealing with overlapping virtual communities in Big data. The knowledge encoding based on fuzzy–rough set provides superior performance than that of rough set. Future directions of research and challenges including the significance of z-numbers in precisiation of granules are stated. The article includes some of the results published elsewhere.

Three-way k-means: integrating k-means and three-way decision

The traditional k-means, which unambiguously assigns an object precisely to a single cluster with crisp boundary, does not adequately show the fact that a cluster may not have a well-defined cluster boundary. This paper presents a three-way k-means clustering algorithm based on three-way strategy. In the proposed method, an overlap clustering is used to obtain the supports (unions of the core regions and the fringe regions) of the clusters and perturbation analysis is applied to separate the core regions from the supports. The difference between the support and the core region is regarded as the fringe region of the specific cluster. Therefore, a three-way explanation of the cluster is naturally formed. Davies–Bouldin index (DB), Average Silhouette index (AS) and Accuracy (ACC) are computed by using core region to evaluate the structure of three-way k-means result. The experimental results on UCI data sets and USPS data sets show that such strategy is effective in improving the structure of clustering results.

A fuzzy rule base approach for High Impedance Fault detection in distribution system using Morphology Gradient filter

This paper presents a Morphology Gradient (MG) filter-based method for detecting High Impedance Faults (HIFs) from non-HIF events. A MG filter is used to extract the statistical features of the current signal. Then, a Random Forest (RF) rule set is created from the statistical features. These rules are with crisp boundary values. Hence, fuzzy membership functions and a fuzzy rule base is developed to discriminate HIF from non-HIF events. From the results of the proposed method, it is found that the method could detect and differentiate HIFs from normal system events with high dependability. Characteristics of the proposed method are analyzed through a broad simulation study of an actual distribution system and found that this method is capable of detecting HIF with negligible effect of variation in operating conditions.

Deep Crisp Boundaries: From Boundaries to Higher-Level Tasks.

Edge detection has made significant progress with the help of deep convolutional networks (ConvNet). These ConvNet-based edge detectors have approached human level performance on standard benchmarks. We provide a systematical study of these detectors' outputs. We show that the detection results did not accurately localize edge pixels, which can be adversarial for tasks that require crisp edge inputs. As a remedy, we propose a novel refinement architecture to address the challenging problem of learning a crisp edge detector using ConvNet. Our method leverages a top-down backward refinement pathway, and progressively increases the resolution of feature maps to generate crisp edges. Our results achieve superior performance, surpassing human accuracy when using standard criteria on BSDS500, and largely outperforming the state-of-the-art methods when using more strict criteria. More importantly, we demonstrate the benefit of crisp edge maps for several important applications in computer vision, including optical flow estimation, object proposal generation, and semantic segmentation.

\u03b4-MAPS: from spatio-temporal data to a weighted and lagged network between functional domains

In real physical systems the underlying spatial components might not have crisp boundaries and their interactions might not be instantaneous. To this end, we propose δ-MAPS; a method that identifies spatially contiguous and possibly overlapping components referred to as domains, and identifies the lagged functional relationships between them. Informally, a domain is a spatially contiguous region that somehow participates in the same dynamic effect or function. The latter will result in highly correlated temporal activity between grid cells of the same domain. δ-MAPS first identifies the epicenters of activity of a domain. Next, it identifies a domain as the maximum possible set of spatially contiguous grid cells that include the detected epicenters and satisfy a homogeneity constraint. After identifying the domains, δ-MAPS infers a functional network between them. The proposed network inference method examines the statistical significance of each lagged correlation between two domains, applies a multiple-testing process to control the rate of false positives, infers a range of potential lag values for each edge, and assigns a weight to each edge reflecting the magnitude of interaction between two domains. δ-MAPS is related to clustering, multivariate statistical techniques and network community detection. However, as we discuss and also show with synthetic data, it is also significantly different, avoiding many of the known limitations of these methods.We illustrate the application of δ-MAPS on data from two domains: climate science and neuroscience. First, the sea-surface temperature climate network identifies some well-known teleconnections (such as the lagged connection between the El Nintilde {o} Southern Oscillation and the Indian Ocean). Second, the analysis of resting state fMRI cortical data confirms the presence of known functional resting state networks (default mode, occipital, motor/somatosensory and auditory), and shows that the cortical network includes a backbone of relatively few regions that are densely interconnected.

Acoustic shape optimization using cut finite elements

SummaryFictitious domain methods are attractive for shape optimization applications, since they do not require deformed or regenerated meshes. A recently developed such method is the CutFEM approach, which allows crisp boundary representations and for which uniformly well‐conditioned system matrices can be guaranteed. Here, we investigate the use of the CutFEM approach for acoustic shape optimization, using as test problem the design of an acoustic horn for favorable impedance‐matching properties. The CutFEM approach is used to solve the Helmholtz equation, and the geometry of the horn is implicitly described by a level‐set function. To promote smooth algorithmic updates of the geometry, we propose to use the nodal values of the Laplacian of the level‐set function as design variables. This strategy also improves the algorithm's convergence rate, counteracts mesh dependence, and, in combination with Tikhonov regularization, controls small details in the optimized designs. An advantage with the proposed method is that the exact derivatives of the discrete objective function can be expressed as boundary integrals, as opposed to when using a traditional method that uses mesh deformations. The resulting horns possess excellent impedance‐matching properties and exhibit surprising subwavelength structures, not previously seen, which are possible to capture due to the fixed mesh approach.

Fuzzy competence model drift detection for data-driven decision support systems

This paper focuses on concept drift in business intelligence and data-driven decision support systems (DSSs). The assumption of a fixed distribution in the data renders conventional static DSSs inaccurate and unable to make correct decisions when concept drift occurs. However, it is important to know when, how, and where concept drift occurs so a DSS can adjust its decision processing knowledge to adapt to an ever-changing environment at the appropriate time. This paper presents a data distribution-based concept drift detection method called fuzzy competence model drift detection (FCM-DD). By introducing fuzzy sets theory and replacing crisp boundaries with fuzzy ones, we have improved the competence model to provide a better, more refined empirical distribution of the data stream. FCM-DD requires no prior knowledge of the underlying distribution and provides statistical guarantee of the reliability of the detected drift, based on the theory of bootstrapping. A series of experiments show that our proposed FCM-DD method can detect drift more accurately, has good sensitivity, and is robust.

Sensitivity Association Rule Mining using Weight based Fuzzy Logic

Mining of sensitive rules is the most important task in data mining. Most of the existing techniques worked on finding sensitive rules based upon the crisp thresh hold value of support and confidence which cause serious side effects to the original database. To avoid these crisp boundaries this paper aims to use WFPPM (Weighted Fuzzy Privacy Preserving Mining) to extract sensitive association rules. WFPPM completely find the sensitive rules by calculating the weights of the rules. At first, we apply FP-Growth to mine association rules from the database. Next, we implement fuzzy to find the sensitive rules among the extracted rules. Experimental results show that the proposed scheme find actual sensitive rules without any modification along with maintaining the quality of the released data as compared to the previous techniques.

Neutrosophic Crisp Mathematical Morphology

In this paper, we aim to apply the concepts of the neutrosophic crisp sets and its operations to the classical mathematical morphological operations, introducing what wecall Neutrosophic Crisp Mathematical Morphology. Sever-al operators are to be developed, including the neutrosophic crisp dilation, the neutrosophic crisp erosion, the neutrosoph-ic crisp opening and the neutrosophic crisp closing. Moreover, we extend the definition of some morphological filters using the neutrosophic crisp sets concept. For instance, we introduce the neutrosophic crisp boundary extraction, the neutrosophic crisp Top-hat and the neutrosophic crisp Bot-tom-hat filters. The idea behind the new introduced operators and filters is to act on the image in the neutrosophic crisp domain instead of the spatial domain.

Fuzzy V(variation)-level clustering

Clustering is one of the important techniques in data mining as well as one of the most fundamental issues in pattern recognition and business intelligence methods. Traditional conventional methods and evolutionary bio-inspired methods are applied for detecting hard clusters. In most of the real time problems, identification and analysis of hard clusters is impossible or may not be suitable for further analysis and prediction. It leads the requirement for soft clustering identification using fuzzy. The paper proposes about one such fuzzy soft clustering model using Vvariation-level to remove sharp crisp boundaries between clusters. The proposed V-level fuzzy clustering algorithm is capable of handling mixed type data items as well. Experiments were done with both deterministic crisp algorithms, and fuzzy algorithms and the results were compared to observe the efficiency of the model to form soft clusters.

Evaluation of Deterministic and Complex Analytical Hierarchy Process Methods for Agricultural Land Suitability Analysis in a Changing Climate

Land suitability analysis is employed to evaluate the appropriateness of land for a particular purpose whilst integrating both qualitative and quantitative inputs, which can be continuous in nature. However, in agricultural modelling there is often a disregard of this contiguous aspect. Therefore, some parametric procedures for suitability analysis compartmentalise units into defined membership classes. This imposition of crisp boundaries neglects the continuous formations found throughout nature and overlooks differences and inherent uncertainties found in the modelling. This research will compare two approaches to suitability analysis over three differing methods. The primary approach will use an Analytical Hierarchy Process (AHP), while the other approach will use a Fuzzy AHP over two methods; Fitted Fuzzy AHP and Nested Fuzzy AHP. Secondary to this, each method will be assessed into how it behaves in a climate change scenario to understand and highlight the role of uncertainties in model conceptualisation and structure. Outputs and comparisons between each method, in relation to area, proportion of membership classes and spatial representation, showed that fuzzy modelling techniques detailed a more robust and continuous output. In particular the Nested Fuzzy AHP was concluded to be more pertinent, as it incorporated complex modelling techniques, as well as the initial AHP framework. Through this comparison and assessment of model behaviour, an evaluation of each methods predictive capacity and relevance for decision-making purposes in agricultural applications is gained.

The uncertainty of probabilistic rough sets in multi-granulation spaces

Pawlak's rough sets model describes an uncertain target set (concept) with two crisp boundary lines (i.e. lower and upper approximation sets) and as an effective tool has successfully been used to deal with uncertain information systems. Based on the rough sets model, a probabilistic rough sets model with a pair of thresholds was proposed to improve the fault-tolerance ability of rough sets. The uncertainty of Pawlak's rough sets model is rooted in the objects contained in the boundary region of the target concept, while the uncertainty of probabilistic rough sets model comes from three regions, because the objects in the positive or negative regions are probably uncertain, and the membership degrees of these objects are not necessary equal to 1 or 0. In this paper, a method for measuring the uncertainty of probabilistic rough sets is proposed, and the change rules of uncertainty with changing knowledge spaces are presented and analyzed. Then, for an uncertain target concept, the uncertainties of the three regions are discussed, and the related change rules of uncertainty with changing knowledge spaces are revealed and successfully proved. Finally, a comparative analysis on the uncertainty of a target concept in rough sets and probabilistic rough sets model is presented. These results are important to further enrich and improve probabilistic rough sets theory, and effectively promote the development of uncertainty artificial intelligence.