Designing efficient deep learning models for 3D point clouds is an important research topic. Point-voxel convolution (Liu et al. in NeurIPS, 2019) is a pioneering approach in this direction, but it ...still has considerable room for improvement in terms of performance, since it has quite a few layers of simple 3D convolutions and linear point-voxel feature fusion operations. To resolve these issues, we propose a novel reparameterizable point-voxel convolution (RepPVConv) block. First, RepPVConv adopts two reparameterizable 3D convolution modules to extract more informative voxel features without introducing any extra computational overhead for inference. The rationale is that the reparameterizable 3D convolution modules are trained in high-capacity modes but are reparameterized into low-capacity modes during inference while losslessly maintaining the original performance. Second, RepPVConv attentively fuses the reparameterized voxel features with those of points. Since the proposed approach operates in a nonlinear manner, descriptive reparameterized voxel features can be better utilized. Extensive experimental results show that RepPVConv-based networks are efficient in terms of both GPU memory consumption and computational complexity and significantly outperform the state-of-the-art methods.
3D face reconstruction has witnessed considerable progress in recovering 3D face shapes and textures from in-the-wild images. However, due to a lack of texture detail information, the reconstructed ...shape and texture based on deep learning could not be used to re-render a photorealistic facial image since it does not work in harmony with weak supervision only from the spatial domain. In the paper, we propose a method of spatio-frequency decoupled weak-supervision for face reconstruction, which applies the losses from not only the spatial domain but also the frequency domain to learn the reconstruction process that approaches photorealistic effect based on the output shape and texture. In detail, the spatial domain losses cover image-level and perceptual-level supervision. Moreover, the frequency domain information is separated from the input and rendered images, respectively, and is then used to build the frequency-based loss. In particular, we devise a spectrum-wise weighted Wing loss to implement balanced attention on different spectrums. Through the spatio-frequency decoupled weak-supervision, the reconstruction process can be learned in harmony and generate detailed texture and high-quality shape only with labels of landmarks. The experiments on several benchmarks show that our method can generate high-quality results and outperform state-of-the-art methods in qualitative and quantitative comparisons.
•Based on the assumption of equivalent energy, a hybrid of FEA and top-down CG model is successfully formulated and generalized to a variety of cases.•CMA-ES algorithm turns out to be an effective ...tool to determine the best-fitting parameters for the Morse bond potential.•The proposed approach is adaptive in describing fracture behavior of 2D metamaterials, generating much smoother centered notch boundary, which geometrically agrees with the all-atom model.
Motivated by the demand of computational flexibility and efficiency, we present a novel approach for coarse-grained (CG) modeling of notched 2D metamaterials, highlighting a “top-down” strategy with adaptive triangular mesh and Morse bond potential. The proposed approach focuses on a typical 2D material of graphene with different circular notches as a case study. Particularly, based on the assumption of equivalent energy, a triangular mesh is adopted to map areas into triangles, which is optimized by using the covariance matrix adaptation evolution strategy (CMA-ES) algorithm to determine the best-fitting parameters of the Morse bond potential, for models with both uniform and non-uniform meshes. Instead of relying on theoretical calculations, an exponential relationship is formulated between the Morse bond potential parameters and the bond length. The numerical results show that the CMA-ES algorithm demonstrates excellent convergence for all bond lengths, with models taking approximately 20 generations to converge. Compared with conventional bottom-up coarse-grained modeling schemes, the proposed approach leads to much smoother boundaries around notches and displays decent adaptability in a variety of models. The results show that the proposed model is highly consistent with the full-atom model, and is proved to be effective in studying the mechanical properties of notched 2D metamaterials, paving the way for the design and development of advanced 2D metamaterials with inherent or prefabricated notches.
In fusion welded joints fatigue crack originates from the weld toe or weld route due to geometrical discontinuities inducing high stress concentration. This holds true even in presence of residual ...stress which was found to follow the same asymptotic distribution theorized first by Williams and then, in the elastic-plastic filed, by Hutchinson, Rice and Rosengren. Solid state phase transformations play a fundamental rule in determining the sign of the singular residual stress field. This obviously will influence the crack initiation life in the high cycle fatigue regime where plastic effects at the weld toe can be neglected. In this scenario, the influence of process parameters on the residual asymptotic stress distribution is not yet investigated using a 3D numerical modeling of the welding process. This work is aimed at filling this gap. Computational welding mechanics and the peak stress method are combined to calculate the residual Notch Stress Intensity Factors (R-NSIF) as a function of power input. Results are summarized in terms of heat input and discussed in view of a future guidelines proposal addressed to designers who are going to face a welding process optimization for fatigue life improvement.
This paper is focused on the formulation of a numerical model for dynamic crack propagation on concrete aggregate interface. The concept of energy release rate is incorporated in the algorithm that ...is conducted in Abaqus through Python script interface. The proposed model is capable of manipulating free propagation of interface cracks. Thus, the on-surface growth of the interface crack, as well as the crack penetration into the concrete matrix, are successfully implemented. A case in point is the rupture behavior of concrete matrix containing one-single aggregate. Simulation of one matrix containing an isolated aggregate was conducted. Influence of the side-edge constraint, the aggregate direction as well as the fracture energy of the interface, was investigated. The results show that, tensile constraint on the side edge, a smaller angle between tensile axis and aggregate, and higher fracture energy could lead to a higher rupture strength of the interface. Once the interface starts to grow, it immediately and unstably propagates to the two ends of the aggregate major axis, and further enters the matrix. The three factor influences less on the character of above rupture path. Though the conclusion is prudently stipulated to concrete matrix with single aggregate, the numerical model can be also further modified to study the trans-scale propagations of multiple cracks in concrete materials or components.
The behaviors of short fatigue crack remain to be one of the most challenging topics in fracture mechanics. It is widely accepted that the process of short fatigue crack can be divided as ...microstructure‐sensitive stage and microstructure‐independent stage. The inherent multiscaling characteristics necessitates the formulation of multiscale short fatigue crack growth model. One crucial way to address the concern is to explore physically reasonable expression of crack‐tip driving force. This paper briefly reviews the short fatigue crack growth models through two approaches. One way is the further modification of conventional stress intensity factor (SIF) range. The other way is to seek the replacement of SIF range, by employing J‐integral and strain energy density factor. Reflections on the present multiscale models are summarized in the concluding remarks.
Although data-driven models, especially deep learning, have achieved astonishing results on many prediction tasks for nonlinear sequences, challenges still remain in finding an appropriate way to ...embed prior knowledge of physical dynamics in these models. In this work, we introduce a convex relaxation approach for learning robust Koopman operators of nonlinear dynamical systems, which are intended to construct approximate space spanned by eigenfunctions of the Koopman operator. Different from the classical dynamic mode decomposition, we use the layer weights of neural networks as eigenfunctions of the Koopman operator, providing intrinsic coordinates that globally linearize the dynamics. We find that the approximation of space can be regarded as an orthogonal Procrustes problem on the Stiefel manifold, which is highly sensitive to noise. The key contribution of this paper is to demonstrate that strict orthogonal constraint can be replaced by its convex relaxation, and the performance of the model can be improved without increasing the complexity when dealing with both clean and noisy data. After that, the overall model can be optimized via backpropagation in an end-to-end manner. The comparisons of the proposed method against several state-of-the-art competitors are shown on nonlinear oscillators and the lid-driven cavity flow.
Designing efficient deep learning models for 3D point cloud perception is becoming a major research direction. Point-voxel convolution (PVConv) Liu et al. (2019) is a pioneering research work in this ...topic. However, since with quite a few layers of simple 3D convolutions and linear point-voxel feature fusion operations, it still has considerable room for improvement in performance. In this paper, we propose a novel pyramid point-voxel convolution (PyraPVConv) block with two key structural modifications to address the above issues. First, PyraPVConv uses a voxel pyramid module to fully extract voxel features in the manner of feature pyramid, such that sufficient voxel features can be obtained efficiently. Second, a sharable attention module is utilized to capture compatible features between multi-scale voxels in pyramid and point cloud for aggregation, as well as to reduce the complexity via structure sharing. Extensive results on three point cloud perception tasks, i.e., indoor scene segmentation, object part segmentation and 3D object detection, validate that the networks constructed by stacking PyraPVConv blocks are efficient in terms of both GPU memory consumption and computational complexity, and are superior to the state-of-the-art methods.
Adversarial attacks have been successfully extended to the field of point clouds. Besides applying the common perturbation guided by the gradient, adversarial attacks on point clouds can be conducted ...by applying directional perturbations, e.g., along normal and along the tangent plane. In this article, we first investigate whether adversarial attacks with these two orthogonal directional perturbations are more imperceptible than that with the gradient-aware perturbation. Second, we investigate the deeper difference between adversarial attacks with these two directional perturbations, and whether they are applicable to the same scenarios. Third, based on the verification results that the above two directional perturbations have different sensitiveness to curvature, we devise a novel normal-tangent attack (NTA) framework with a hybrid directional perturbation scheme that adaptively chooses the direction according to the curvature of the local shape around the point. Extensive experiments on two publicly available data sets, e.g., ModelNet40 and ShapeNet Part, with classifiers in three representative networks, e.g., PointNet++, DGCNN, PointConv, validate the effectiveness of NTA, and the superiority to the state-of-the-art methods.
PDF
