Search

Deep neural networks are known to be vulnerable to adversarial attacks. The empirical analysis in our study suggests that attacks tend to induce diverse network architectures to shift the attention to irrelevant regions. Motivated by this observation, we propose a regularization technique which enforces the attentions to be well aligned via the knowledge transfer mechanism, thereby encouraging the robustness. Resultant model exhibits unprecedented robustness, securing 63.81 % adversarial accuracy where the prior art is 51.59 % on CIFAR-10 dataset under PGD attacks. In addition, we go beyond performance to analytically investigate the proposed method as an effective defense. Significantly flattened loss landscape can be observed, demonstrating the promise of the proposed method for improving robustness and thus the deployment in security-sensitive settings.

MicroRNA-21 (miR-21), probably one of the most studied miRNAs to date, is found pleiotropic in various biological events. Its emerging role in pulmonary remodeling has attracted extensive attention. This review summarizes the genomic information of its primary transcript and various transcriptional regulations on its promoter. In addition, the role of miR-21 in pulmonary remodeling related signaling such as transforming growth factor β (TGF-β), bone morphogenetic protein (BMP), epidermal growth factor receptor (EGFR) and Notch signaling is discussed. Various validated miR-21 target genes participate in controlling of the overactive cell accumulation, smooth muscle contraction, inflammatory stress (trigger for lung epithelium damage), extracellular matrix deposition and hypoxia-induced disorders. Moreover, we focus on its particular implication in events including inflammatory stress-driven epithelium damage, epithelial-to-mesenchymal transition (EMT), transdifferentiation of fibroblasts into myofibroblasts, hypoxia stimuli and ROS response, as well as some other pulmonary remodeling related events such as overactive fibroblast (myofibroblast) accumulation, extracellular matrix deposition, and angiogenesis. Here, we summarize the strong potential of miR-21 in pulmonary remodeling and provide novel clues for further research in this area.

Principal component analysis (PCA), which is also known as Karhunen-Loeve (KL) transform, is a classical statistic technique that has been applied to many fields, such as knowledge representation, pattern recognition and image compression. The objective of PCA is to reduce the dimensionality of dataset and identify new meaningful underlying variables. The key idea is to project the objects to an orthogonal subspace for their compact representations. It usually involves a mathematical procedure that transforms a number of correlated variables into a smaller number of uncorrelated variables, which are called principal components. The first principal component accounts for as much of the variability in the dataset as possible, and each succeeding component accounts for as much of the remaining variability as possible. In pattern recognition, PCA technique was first applied to the representation of human face images by Sirovich and Kirby in [1,2]. This then led to the well-known Eigenfaces method for face recognition proposed by Turk and Penland in [3]. Since then, there has been an extensive literature that addresses both the theoretical aspect of the Eigenfaces method and its application aspect [4-6]. In image compression, PCA technique has also been widely applied to the remote hyperspectral imagery for classification and compression [7,8]. Nevertheless, it can be noted that in the classical 1DPCA scheme the 2D data sample (e.g. image) must be initially converted to a 1D vector form. The resulting sample vector will lead to a high dimensional vector space. It is consequently difficult to evaluate the covariance matrix accurately when the sample vector is very long and the number of training samples is small. Furthermore, it can also be noted that the projection of a sample on each principal orthogonal vector is a scale. Obviously, this will cause the sample data to be over-compressed. In order to solve this kind of dimensionality problem, Yang et al. [9,10] proposed the 2D-PCA approach. The basic idea is to directly use a set of matrices to construct the corresponding covariance matrix instead of a set of vectors. Compared with the covariance matrix of 1D-PCA, one can note that the size of the covariance matrix using 2D-PCA is much smaller. This improves the computational efficiency. Furthermore, it can be noted that the projection of a sample on each principal orthogonal vector is a vector. Thus, the problem of over-compression is alleviated in the 2DPCA scheme. In addition, Wang et al. [11] proposed that the 2D-PCA was equivalent to a special case of the block-based PCA, and emphasized that this kind of block-based methods had been used for face recognition in a number of systems. 2

To deal with data sets from real-time 3D sensors of RGB-D or TOF cameras, this paper presents a method for registration of unstructured point clouds. We firstly derive intrinsic shape context descriptors for 3D data organization. To replace the Fast-Marching method, a vertex-oriented triangle propagation method is applied to calculate the ’angle’ and ’radius’ in descriptor charting, so that the matching accuracy at the twisting and folding area is significantly improved. Then, a 3D cylindrical shape descriptor is proposed for registration of unstructured point clouds. The chosen points are projected into the cylindrical coordinate system to construct the descriptors. The projection parameters are respectively determined by the distances from the chosen points to the reference normal vector, and the distances from the chosen points to the reference tangent plane and the projection angle. Furthermore, Fourier transform is adopted to deal with orientation ambiguity in descriptor matching. Practical experiments demonstrate a satisfactory result in point cloud registration and notable improvement on standard benchmarks.

Long-term stability and selectivity are crucial for the practical application of gas sensors, which are closely related to the microstructure and composition of sensor materials. In this work, Co-based metal organic framework (MOF) is used as a precursor and prepared by a simple hydrothermal method. After calcination, a series of Co3O4 nanospheres with various microstructures are derived. When the calcination temperature increases from 300 °C to 500 °C, the microstructure of Co3O4 nanospheres changed from rough solid to porous, and then transformed into porous core-shell. When assembled into the gas sensors, the Co3O4 nanospheres with porous structure calcined at 400 °C (Co3O4-400) show the highly selective response of 53.78 for 100 ppm n-butanol at the operating temperature of 140 °C. Moreover, the theoretical limit of detection was calculated to be 150 ppb. The reproducibility, selectivity and stability of the gas sensor were further verified to be excellent. After 45 days, the response value of Co3O4-400 is at 86.74%, even after 75 days, the response value remains at 74.93%. The main reason can be attributed to the large specific surface area, abundant pore structure and a large number of oxygen vacancies on its surface. These findings provide reference for the development of p-type metal oxide semiconductor (MOS) sensors with long-term stability and high performance.

OAPA Benefiting from the joint consideration of geometric structures and low-rank constraint, graph low-rank representation (GLRR) method has led to the state-of-the-art results in many applications. However, it faces the limitations that the structure of errors should be known a prior, the isolated construction of graph Laplacian matrix, and the over shrinkage of the leading rank components. To improve GLRR in these regards, this paper proposes a new LRR model, namely iterative reconstrained LRR via weighted nonconvex regularization (IRWNR), using three distinguished properties on the concerned representation matrix. The first characterizes various distributions of the errors into an adaptively learned weight factor for more flexibility of noise suppression. The second generates an accurate graph matrix from weighted observations for less afflicted by noisy features. The third employs a parameterized Rational function to reveal the importance of different rank components for better approximation to the intrinsic subspace structure. Following a deep exploration of automatic thresholding, parallel update, and partial SVD operation, we derive a computationally efficient low-rank representation algorithm using an iterative reconstrained framework and accelerated proximal gradient method. Comprehensive experiments are conducted on synthetic data, image clustering, and background subtraction to achieve several quantitative benchmarks as clustering accuracy, normalized mutual information, and execution time. Results demonstrate the robustness and efficiency of IRWNR compared with other state-of-the-art models.

Relief is an art form part way between 3D sculpture and 2D painting. We present a novel approach for generating a texture-mapped high-relief model from a single brush painting. Our aim is to extract the brushstrokes from a painting and generate the individual corresponding relief proxies rather than recovering the exact depth map from the painting, which is a tricky computer vision problem, requiring assumptions that are rarely satisfied. The relief proxies of brushstrokes are then combined together to form a 2.5D high-relief model. To extract brushstrokes from 2D paintings, we apply layer decomposition and stroke segmentation by imposing boundary constraints. The segmented brushstrokes preserve the style of the input painting. By inflation and a displacement map of each brushstroke, the features of brushstrokes are preserved by the resultant high-relief model of the painting. We demonstrate that our approach is able to produce convincing high-reliefs from a variety of paintings(with humans, animals, flowers, etc.). As a secondary application, we show how our brushstroke extraction algorithm could be used for image editing. As a result, our brushstroke extraction algorithm is specifically geared towards paintings with each brushstroke drawn very purposefully, such as Chinese paintings, Rosemailing paintings, etc.

Subspace clustering aims to partition the data points\ud drawn from a union of subspaces according to their underly-\ud ing subspaces. For accurate semisupervised subspace clustering,all data that have a must-link constraint or the same label should be grouped into the same underlying subspace. However, this is not guaranteed in existing approaches. Moreover, these approaches require additional parameters for incorporating supervision information. In this paper, we propose a constrained low-rank representation (CLRR) for robust semisupervised sub-space clustering, based on a novel constraint matrix constructed\ud in this paper. While seeking the low-rank representation of data, CLRR explicitly incorporates supervision information as hard constraints for enhancing the discriminating power of optimal representation. This strategy can be further extended to other state-of-the-art methods, such as sparse subspace clustering. We theoretically prove that the optimal representation matrix has both a block-diagonal structure with clean data and a semisu-pervised grouping effect with noisy data. We have also developed an efficient optimization algorithm based on alternating the direction method of multipliers for CLRR. Our experimental results have demonstrated that CLRR outperforms existing methods.

LiFe0.5Mn0.5PO4 typically exhibits two pairs of plateaus in charge/discharge curves associated with MnPO4/LiMnPO4 and FePO4/LiFePO4 phase transformation. In this work, an additional plateau was observed at about 3.7 V in the discharge curve at relative high current rates, but not in the charge process. The additional plateau is identified to Mn3+ reduction by capacity analysis, cyclic voltammograms (CV) and galvanostatic intermittent titration technique (GITT) measurement. The capacity of additional plateau increases with the increase of discharge rates. LiFe0.5Mn0.5PO4 discharge mainly proceeds at this plateau at high current rate. Keywords: LiFe0.5Mn0.5PO4, Additional discharge plateau, Cathode material, Lithium ions battery

© 2017 IEEE.Southeast Asia is one of the most rapidly growing regions in the world with natural and cultural resources. It is important to pass on the cultural knowledge to the next generation. Intangible Cultural Heritage like traditional dances, or folk dance is a valuable cultural knowledge to be maintained and passed on by transferring tacit knowledge, and even explicit knowledge such as books, or video presentations. Issues of passing on the knowledge can be the loss of knowledge from time to time by the reduction of the number of dance masters, unreliable sources, and low quality. To retrieve such valuable knowledge, there is a widely-used tool in Europe, in the United States, Asia and Southeast Asia, called 'Labanotation' which is about recording and analyzing the dance movement. This paper focuses on proposing a framework for a traditional Thai dance knowledge archive creating an ontology using knowledge engineering based on Labanotation by transferring notation scores to represent the dance in 3D Animation. The framework assists dancers, notators, knowledge engineers, software engineers to successfully communicate with each other.

Multi-view clustering, which explores complementary \ud information between multiple distinct feature sets, has received considerable attention. For accurate clustering, all data with the same label should be clustered together regardless of their multiple views. However, this is not guaranteed in existing approaches. To address this issue, we propose Adaptive Multi-View Semi-Supervised Nonnegative Matrix Factorization (AMVNMF), which uses label information as hard constraints to ensure data with same label are clustered together, so that the discriminating power of new representations are enhanced. Besides, AMVNMF provides a viable solution to learn the weight of each view adaptively\ud with only a single parameter. Using L2,1-norm, AMVNMF is also robust to noises and outliers. We further develop an efficient iterative algorithm for solving the optimization problem. Experiments carried out on five well-known datasets have demonstrated the effectiveness of AMVNMF\ud in comparison to other existing state-of-the-art approaches in terms of accuracy and normalized mutual information.

Computing discrete geodesic distance over triangle meshes is one of the fundamental problems in computational geometry and computer graphics. In this problem, an effective window pruning strategy can significantly affect the actual running time. Due to its importance, we conduct an in-depth study of window pruning operations in this paper, and produce an exhaustive list of scenarios where one window can make another window partially or completely redundant. To identify a maximal number of redundant windows using such pairwise cross checking, we propose a set of procedures to synchronize local window propagation within the same triangle by simultaneously propagating a collection of windows from one triangle edge to its two opposite edges. On the basis of such synchronized window propagation, we design a new geodesic computation algorithm based on a triangle-oriented region growing scheme. Our geodesic algorithm can remove most of the redundant windows at the earliest possible stage, thus significantly reducing computational cost and memory usage at later stages. In addition, by adopting triangles instead of windows as the primitive in propagation management, our algorithm significantly cuts down the data management overhead. As a result, it runs 4--15 times faster than MMP and ICH algorithms, 2-4 times faster than FWP-MMP and FWP-CH algorithms, and also incurs the least memory usage.

Matrilin-1 (Matn1), a cartilage-specific peri-cellular and extracellular matrix (ECM) protein, has been hypothesized to regulate ECM interactions and transmit mechanical signals in cartilage. Since Matn1 knock-out (Matn1-/-) mice exhibit a normal skeleton, its function in vivo is unclear. In this study, we found that the anabolic Acan and Col2a transcript levels were significantly higher in wildtype (Matn1+/+) mouse cartilage than that of MATN1-/- mice in vivo. However, such difference was not observed between Matn1+/+ and MATN1-/- chondrocytes cultured under stationary conditions in vitro. Cyclic loading significantly stimulated Acan and Col2a transcript levels in Matn1+/+ but not in MATN1-/- chondrocytes. This suggests that, while Matn1+/+ chondrocytes increase their anabolic gene expression in response to mechanical loading, the MATN1-/- chondrocytes fail to do so because of the deficiency in mechanotransduction. We also found that altered elastic modulus of cartilage matrix in Matn1-/- mice, suggesting the mechanotransduction has changed due to the deficiency of Matn1. To understand the impact of such deficiency on joint disease, mechanical loading was altered in vivo by destabilization of medial meniscus. While Matn1+/+ mice exhibited superficial fissures and clefts consistent with mechanical damage to the articular joint, Matn1-/- mice presented more severe cartilage lesions characterized by proteoglycan loss and disorganization of cells and ECM. This suggests that Matn1 deficiency affects pathogenesis of post-traumatic osteoarthritis by failing to up-regulate anabolic gene expression. This is the first demonstration of Matn1 function in vivo, which suggests its protective role in cartilage degeneration under altered mechanical environment.

We present a novel method for caricature synthesis based on mean value coordinates (MVC). Our method can be applied to any single frontal face image to learn a specified caricature face pair for frontal and 3D caricature synthesis. This technique only requires one or a small number of exemplar pairs and a natural frontal face image training set, while the system can transfer the style of the exemplar pair across individuals. Further exaggeration can be fulfilled in a controllable way. Our method is further applied to facial expression transfer, interpolation, and exaggeration, which are applications of expression editing. Additionally, we have extended our approach to 3D caricature synthesis based on the 3D version of MVC. With experiments we demonstrate that the transferred expressions are credible and the resulting caricatures can be characterized and recognized.

Texture mapping has long been used in computer graphics to \ud enhance the realism of virtual scenes. However, to match the 3D model feature points with the corresponding pixels in a texture image, surface parameterization must satisfy specific positional constraints. However, despite numerous \ud research efforts, the construction of a mathematically robust, foldover‐free parameterization that is subject to \ud positional constraints continues to be a challenge. In the \ud present paper, this foldover problem is addressed by developing radial basis function (RBF) based reparameterization. Given initial 2D embedding of a 3D \ud surface, the proposed method can reparameterize 2D embedding into a foldover ‐free 2D mesh, satisfying a set \ud of user‐specified constraint points. In addition, this approach is mesh‐free. Therefore, generating smooth texture \ud mapping results is possible without extra smoothing optimization.

The likeness of a synthesized caricature and the original face image is an essential and often overlooked part of caricature production. In this paper, we present an example based caricature synthesis technique, consisting of shape exaggeration, relationship exaggeration, and optimization for likeness. Rather than relying on a large training set of caricature face pairs, our shape exaggeration step is based on only one or a small number of examples of facial elements (e.g., eyes and nose). The relationship exaggeration step introduces two definitions which facilitate global facial feature synthesis. The first is the T-shape rule, which describes the relative relationship between the facial elements in an intuitive manner. The second is the so-called proportions, which characterizes the facial features in a proportion form. Finally, we introduce a similarity metric as the likeness metric based on the Modified Hausdorff Distance (MHD) which allows us to optimize the configuration of facial elements, maximizing the likeness while satisfying a number of constraints. The effectiveness of our algorithm is demonstrated with experimental results.

In this paper, the min/max flow scheme for image restoration is revised. The novelty consists of the fol-\ud 24 lowing three parts. The first is to analyze the reason of the speckle generation and then to modify the\ud 25 original scheme. The second is to point out that the continued application of this scheme cannot result\ud 26 in an adaptive stopping of the curvature flow. This is followed by modifications of the original scheme\ud 27 through the introduction of the Gradient Vector Flow (GVF) field and the zero-crossing detector, so as\ud 28 to control the smoothing effect. Our experimental results with image restoration show that the proposed\ud 29 schemes can reach a steady state solution while preserving the essential structures of objects. The third is\ud 30 to extend the min/max flow scheme to deal with the boundary leaking problem, which is indeed an\ud 31 intrinsic shortcoming of the familiar geodesic active contour model. The min/max flow framework pro-\ud 32 vides us with an effective way to approximate the optimal solution. From an implementation point of\ud 33 view, this extended scheme makes the speed function simpler and more flexible. The experimental\ud 34 results of segmentation and region tracking show that the boundary leaking problem can be effectively\ud 35 suppressed.

In this paper, the gradient vector flow fields are introduced in image restoration. Within the context of flow fields, the shock filter, mean curvature flow, and Perona-Malik equation are reformulated. Many advantages over the original models can be obtained; these include numerical stability, large capture range, and high-order derivative estimation. In addition, a fairing process is introduced in the anisotropic diffusion, which contains a fourth-order derivative and is reformulated as the intrinsic Laplacian of curvature under the level set framework. By applying this fairing process, the shape boundaries will become more apparent. In order to overcome numerical errors, the intrinsic Laplacian of curvature is computed from the gradient vector flow fields instead of the observed images.

The weighted nuclear norm minimization and truncated nuclear norm minimization are two well-known low-rank constraint for visual applications. In this paper, by integrating their advantages into a unified formulation, we find a better weighting strategy, namely truncated reweighting norm minimization (TRNM), which provides better approximation to the target rank for some specific task. Albeit nonconvex and truncated, we prove that TRNM is equivalent to certain weighted quadratic programming problems, whose global optimum can be accessed by the newly presented reweighting singular value thresholding operator. More importantly, we design a computationally efficient optimization algorithm, namely momentum update and rank propagation (MURP), for the general TRNM regularized problems. The individual advantages of MURP include, first, reducing iterations through nonmonotonic search, and second, mitigating computational cost by reducing the size of target matrix. Furthermore, the descent property and convergence of MURP are proven. Finally, two practical models, i.e., Matrix Completion Problem via TRNM (MCTRNM) and Space Clustering Model via TRNM (SCTRNM), are presented for visual applications. Extensive experimental results show that our methods achieve better performance, both qualitatively and quantitatively, compared with several state-of-the-art algorithms.