3D Mesh Generation Research Articles

Junction conditions for one-dimensional network hemodynamic model for total cavopulmonary connection using physically informed deep learning technique

Abstract This paper presents a novel methodology utilizing physics-informed neural network (PINN) as a junction condition for a 1D network model of blood flow in total cavopulmonary connection generated by the Fontan procedure. The technique integrates a 3D mesh generation process based on the parameterization of the junction geometry, along with a sophisticated physically regularized neural network architecture. Synthetic datasets are produced using 3D steady Stokes simulations within fixed boundaries. We use a physically informed feedforward neural network that utilizes a physically regularized loss function, which incorporates the principle of mass conservation. Our PINN achieves a tolerance of 6% on the test set. We develop a 1D-PINN multiscale model based on a previously developed method for multiscale 1D–3D simulations. Comparison with 1D–3D Stokes based model and 3D Navier–Stokes based model verifies the 1D-PINN model. In the first and second comparison, the maximum deviations of the averaged pressures and flows do not exceed 1.48% and 12.26%, respectively.

Joint Optimization-Based Texture and Geometry Enhancement Method for Single-Image-Based 3D Content Creation

Recent studies have explored the generation of three-dimensional (3D) meshes from single images. A key challenge in this area is the difficulty of improving both the generalization and detail simultaneously in 3D mesh generation. To address this issue, existing methods utilize fixed-resolution mesh features to train networks for generalization. This approach is capable of generating the overall 3D shape without limitations on object categories. However, the generated shape often exhibits a blurred surface and suffers from suboptimal texture resolution due to the fixed-resolution mesh features. In this study, we propose a joint optimization method that enhances geometry and texture by integrating generalized 3D mesh generation with adjustable mesh resolution. Specifically, we apply an inverse-rendering-based remeshing technique that enables the estimation of complex-shaped mesh estimations without relying on fixed-resolution structures. After remeshing, we enhance the texture to improve the detailed quality of the remeshed mesh via a texture enhancement diffusion model. By separating the tasks of generalization, detailed geometry estimation, and texture enhancement and adapting different target features for each specific network, the proposed joint optimization method effectively addresses the characteristics of individual objects, resulting in increased surface detail and the generation of high-quality textures. Experimental results on the Google Scanned Objects and ShapeNet datasets demonstrate that the proposed method significantly improves the accuracy of 3D geometry and texture estimation, as evaluated by the PSNR, SSIM, LPIPS, and CD metrics.

Creating a 3D Mesh in A‐pose from a Single Image for Character Rigging

AbstractLearning‐based methods for 3D content generation have shown great potential to create 3D characters from text prompts, videos, and images. However, current methods primarily focus on generating static 3D meshes, overlooking the crucial aspect of creating an animatable 3D meshes. Directly using 3D meshes generated by existing methods to create underlying skeletons for animation presents many challenges because the generated mesh might exhibit geometry artifacts or assume arbitrary poses that complicate the subsequent rigging process. This work proposes a new framework for generating a 3D animatable mesh from a single 2D image depicting the character. We do so by enforcing the generated 3D mesh to assume an A‐pose, which can mitigate the geometry artifacts and facilitate the use of existing automatic rigging methods. Our approach aims to leverage the generative power of existing models across modalities without the need for new data or large‐scale training. We evaluate the effectiveness of our framework with qualitative results, as well as ablation studies and quantitative comparisons with existing 3D mesh generation models.

Mesh-controllable multi-level-of-detail text-to-3D generation

Text-to-3D generation is a challenging but significant task and has gained widespread attention. Its capability to rapidly generate 3D digital assets holds huge potential application value in fields such as film, video games, and virtual reality. However, current methods often face several drawbacks, including long generation times, difficulties with the multi-face Janus problem, and issues like chaotic topology and redundant structures during mesh extraction. Additionally, the lack of control over the generated results limits their utility in downstream applications. To address these problems, we propose a novel text-to-3D framework capable of generating meshes with high fidelity and controllability. Our approach can efficiently produce meshes and textures that match the text description and the desired level of detail (LOD) by specifying input text and LOD preferences. This framework consists of two stages. In the coarse stage, 3D Gaussians are employed to accelerate generation speed, and weighted positive and negative prompts from various observation perspectives are used to address the multi-face Janus problem in the generated results. In the refinement stage, mesh vertices and faces are iteratively refined to enhance surface quality and output meshes and textures that meet specified LOD requirements. Compared to the state-of-the-art text-to-3D methods, extensive experiments demonstrate that the proposed method performs better in solving the multi-face Janus problem, enabling the rapid generation of 3D meshes with enhanced prompt adherence. Furthermore, the proposed framework can generate meshes with enhanced topology, offering controllable vertices and faces with textures featuring UV adaptation to achieve multi-level-of-detail(LODs) outputs. Specifically, the proposed method can preserve the output’s relevance to input texts during simplification, making it better suited for mesh editing and rendering efficiency. User studies also indicate that our framework receives higher evaluations compared to other methods.

Physically Informed Deep Learning Technique for Estimating Blood Flow Parameters in Four-Vessel Junction after the Fontan Procedure

This study introduces an innovative approach leveraging physics-informed neural networks (PINNs) for the efficient computation of blood flows at the boundaries of a four-vessel junction formed by a Fontan procedure. The methodology incorporates a 3D mesh generation technique based on the parameterization of the junction’s geometry, coupled with an advanced physically regularized neural network architecture. Synthetic datasets are generated through stationary 3D Navier–Stokes simulations within immobile boundaries, offering a precise alternative to resource-intensive computations. A comparative analysis of standard grid sampling and Latin hypercube sampling data generation methods is conducted, resulting in datasets comprising 1.1×104 and 5×103 samples, respectively. The following two families of feed-forward neural networks (FFNNs) are then compared: the conventional “black-box” approach using mean squared error (MSE) and a physically informed FFNN employing a physically regularized loss function (PRLF), incorporating mass conservation law. The study demonstrates that combining PRLF with Latin hypercube sampling enables the rapid minimization of relative error (RE) when using a smaller dataset, achieving a relative error value of 6% on the test set. This approach offers a viable alternative to resource-intensive simulations, showcasing potential applications in patient-specific 1D network models of hemodynamics.

Accurate and semantic 3D reconstruction of maize leaves

The rapid advances in equipment and technology for crop three-dimensional (3D) data acquisition have enabled the high-throughput and high-quality acquisition of crop 3D data. However, the challenge now lies in comprehending unstructured crop 3D data and extracting phenotypes and knowledge. Extracting semantic feature points from plant leaves and achieving high-precision 3D reconstruction is a prevalent and formidable issue in this field. In this study, we present an accurate method for semantic 3D reconstruction of maize leaves. The algorithm utilizes 3D point cloud data of maize leaves as the initial input. Subsequently, the semantic 3D reconstruction of maize leaves is achieved by means of point cloud preprocessing, initial triangular meshing, hole filling, remeshing, mesh parameterization, two-dimensional (2D) edge detection, 2D semantic point interpolation, 3D semantic point selection, mesh generation, and optimization of feature-point position. A total of 56 leaves from five different maize cultivars were reconstructed utilizing this method. The results indicate that the average distance error index dCDMP between the reconstructed mesh vertices and the input point cloud is less than 0.3 cm2 for all five cultivars, and the mean of the maximum distance error index dHDMP for each cultivar is within the range of 0.5–0.8 cm, indicating it is high-accuracy. The visualization of leaves shows good consistency between the reconstructed mesh and corresponding point cloud. Additionally, the reconstructed 3D mesh of the leaves exhibits enhanced triangular facet consistency. Each vertex can be precisely identified as a leaf vein point, leaf edge point, leaf tip point, or leaf surface point, and includes semantic data such as the position on the leaf blade, which can be directly employed for phenotype estimations. High-precision reconstruction of 3D models of crop organs that include semantic information and cultivar resolution yields significant benefits. This technology can improve the resolution and accuracy of crop morphological and structural research, and it is also crucial for crop phenomics, functional-structural plant modelling, and fusion research between plant phenotyping and crop models.

Light-weight 3D mesh generation networks based on multi-stage and progressive knowledge distillation

Due to the high data dimensionality and the complexity of the problem, existing 3D mesh reconstruction models often require significant computational resources to achieve satisfactory results. While lightweight model based on knowledge distillation has been explored in many fields such as image classification, training a lightweight 3D mesh reconstruction model remains a challenging task. In this paper, we propose a method to learn a lightweight 3D mesh reconstruction network using knowledge distillation. Specifically, we introduce a novel approach called multi-stage and progressive knowledge distillation, which effectively enhances the guidance from the teacher network to the student network, thereby improving reconstruction performance. Additionally, we propose a projection-based spatial feature unpooling method to provide more accurate spatial features for the increased spatial points. Experimental results show that our lightweight 3D mesh reconstruction network has comparable performance to existing complex models while greatly reducing the number of parameters. Specifically, our method achieves its 98.97% accuracy while reducing the number of graph neural network parameters to 71.42% of the teacher network.

A fast dynamic smooth adaptive meshing scheme with applications to compressible flow

We develop a fast-running smooth adaptive meshing (SAM) algorithm for dynamic curvilinear mesh generation, which is based on a fast solution strategy of the time-dependent Monge-Ampère (MA) equation, det⁡∇ψ(x,t)=G∘ψ(x,t). The novelty of our approach is a new so-called perturbation formulation of MA, which constructs the solution map ψ via composition of a sequence of near-identity deformations of a reference mesh. Then, we formulate a new version of the deformation method [21] that results in a simple, fast, and high-order accurate numerical scheme and a dynamic SAM algorithm that is of optimal complexity when applied to time-dependent mesh generation for solutions to hyperbolic systems such as the Euler equations of gas dynamics. We perform a series of challenging 2D and 3D mesh generation experiments for grids with large deformations, and demonstrate that SAM is able to produce smooth meshes comparable to state-of-the-art solvers [22,18], while running approximately 200 times faster. The SAM algorithm is then coupled to a simple Arbitrary Lagrangian Eulerian (ALE) scheme for 2D gas dynamics. Specifically, we implement the C-method [64,65] and develop a new ALE interface tracking algorithm for contact discontinuities. We perform numerical experiments for both the Noh implosion problem as well as a classical Rayleigh-Taylor instability problem. Results confirm that low-resolution simulations using our SAM-ALE algorithm compare favorably with high-resolution uniform mesh runs.

Robo-vision! 3D mesh generation of a scene for a robot for planar and non-planar complex objects.

This paper presents a novel architecture to generate a world model in terms of mesh from a continuous image stream with depth information extracted from a robot's ego-vision camera. We propose two algorithms for planar and non-planar mesh generation. A Cartesian grid-based mesh fitting algorithm is proposed for mesh generation of planar objects. For mesh generation of non-planar objects, we propose a Self Organization Map based algorithm. The proposed algorithm better approaches the boundary and overall shape of the objects compared to State-Of-the-Art (SOA). Extensive experiments done on three public datasets show that our method surpasses SOA both qualitatively and quantitatively.

Adaptive multigrid strategy for geometry optimization of large-scale three dimensional molecular mechanics

In this paper, we present an efficient adaptive multigrid strategy for the geometry optimization of large-scale three dimensional molecular mechanics. The resulting method can achieve significantly reduced complexity by exploiting the intrinsic low-rank property of the material configurations and by combining the state-of-the-art adaptive techniques with the hierarchical structure of multigrid algorithms. To be more precise, we develop a oneway multigrid method with adaptive atomistic/continuum (a/c) coupling, e.g., blended ghost force correction (BGFC) [1] approximations with gradient-based a posteriori error estimators on the coarse levels. We utilize state-of-the-art 3D mesh generation techniques to effectively implement the method. For 3D crystalline defects, such as vacancies, micro-cracks and dislocations, compared with brute-force optimization, complexity with superior rates can be observed numerically, and the strategy has a five-fold acceleration in terms of CPU time for systems with 108 atoms.

HYBRID AERIAL SENSOR DATA AS BASIS FOR A GEOSPATIAL DIGITAL TWIN

Abstract. More and more cities declare themselves to be a smart city or plan to be the same. Smart cities require a solid data source as basis for all further actions and the urban digital twin is the basis on which all information is collected and analysed. The urban digital twin is much more than just a 3D city model, but often this together with GIS data is the starting point for the urban digital twin. The basis of the urban digital twin is formed by geospatial data in the form of the geospatial digital twin. The digital twin hereby acts as a kind of hub into which all relevant and available information is included and analysed. To generate a geospatial digital twin aerial sensors that collect multiple data simultaneously, hybrid sensors, are perfectly suited for this task. In aerial data acquisition a new era started with the introduction of the first real hybrid sensor systems, like the Leica CityMapper-2. Hybrid in this context means the combination of an (oblique) camera system with a topographic LiDAR into an integrated aerial mapping system. By combining these complimentary sub-systems into one system the weaknesses of the one system could be compensated by using the alternative data source. An example is the mapping of low-light urban canyons, where image-based systems mostly produce unreliable results. For an LiDAR sensor the geometrical reconstruction of these areas is straight forward and leads to accurate results. The paper gives a detailed overview over the development and technical characteristics of hybrid sensor systems. The process of data acquisition is discussed and strategies for hybrid urban mapping are proposed. Furthermore, the paper provides insights into the advantage of LiDAR data for the 3D Mesh generation for urban modelling and on the possibilities to generate new products from the combination of the single products with the help of GeoAI. Finally, the use and some use cases of the hybrid sensor data and the derived products in the context of the urban digital twin is discussed and with the infinite loop of data, analysis, and action it is shown, that all data from the urban digital twin can only be a snapshot at a given point in time and the data recording and analysis is a permanent loop.

BlenderPhotonics: an integrated open-source software environment for three-dimensional meshing and photon simulations in complex tissues.

.Significance: Rapid advances in biophotonics techniques require quantitative, model-based computational approaches to obtain functional and structural information from increasingly complex and multiscaled anatomies. The lack of efficient tools to accurately model tissue structures and subsequently perform quantitative multiphysics modeling greatly impedes the clinical translation of these modalities.Aim: Although the mesh-based Monte Carlo (MMC) method expands our capabilities in simulating complex tissues using tetrahedral meshes, the generation of such domains often requires specialized meshing tools, such as Iso2Mesh. Creating a simplified and intuitive interface for tissue anatomical modeling and optical simulations is essential toward making these advanced modeling techniques broadly accessible to the user community.Approach: We responded to the above challenge by combining the powerful, open-source three-dimensional (3D) modeling software, Blender, with state-of-the-art 3D mesh generation and MC simulation tools, utilizing the interactive graphical user interface in Blender as the front-end to allow users to create complex tissue mesh models and subsequently launch MMC light simulations.Results: Here, we present a tutorial to our Python-based Blender add-on—BlenderPhotonics—to interface with Iso2Mesh and MMC, which allows users to create, configure and refine complex simulation domains and run hardware-accelerated 3D light simulations with only a few clicks. We provide a comprehensive introduction to this tool and walk readers through five examples, ranging from simple shapes to sophisticated realistic tissue models.Conclusions: BlenderPhotonics is user friendly and open source, and it leverages the vastly rich ecosystem of Blender. It wraps advanced modeling capabilities within an easy-to-use and interactive interface. The latest software can be downloaded at http://mcx.space/bp.

Computational Fluid Dynamic Study of the Pharyngeal Airway Characteristics Before and After Mandibular Setback Surgery in Patients with Mandibular Prognathism

Mandibular prognathism had been one of the most inconvenienced deformities developed in humans. This type of deformity possesses a protruded mandible jaw, and it can cause inconvenience in talking chewing and affect the aesthetic of the person diagnosed with this problem. However, this problem can be solved through mandibular setback surgery. This surgery will reposition the mandible jaw backward. The main concern of this surgery is, by repositioning the mandible jaw backward, the cross-sectional area of the airway will also be reduced. This condition might induce the iatrogenic of Obstructive Sleep Apnea (OSA). Thus, this study analyzed the characteristic of the pharyngeal airway in pre-treatment and post-virtual-operative conditions. This analysis simulates the pharyngeal airway with the Computational Fluid Dynamic (CFD) method. The respondent's data was obtained from computed tomography (CT) scan and was modeled into a 3-dimensional (3D) model. The simulation process takes place using the 3D model, mesh generation, and boundary conditions set-up. The flow pattern, pressure drop, airway wall shear stress, and turbulent kinetic energy were compared between both simulations. The simulations result present that, after the surgery, the pharyngeal airway decreases significantly. The pressure drop, flow pattern, airway wall shear stress (WSS), and turbulent kinetic energy (TKE) seem more significant around the critical plane in the post-treatment condition. However, these factors did not contribute to the OSA as it is not large enough to become the cause of OSA.

A Comparative Study between Scanning Devices for 3D Printing of Personalized Ostomy Patches.

This papers presents a comparative study of three different 3D scanning modalities to acquire 3D meshes of stoma barrier rings from ostomized patients. Computerized Tomography and Structured light scanning methods were the digitization technologies studied in this research. Among the Structured Light systems, the Go!Scan 20 and the Structure Sensor were chosen as the handheld 3D scanners. Nineteen ostomized patients took part in this study, starting from the 3D scans acquisition until the printed ostomy patches validation. 3D mesh processing, mesh generation and 3D mesh comparison was carried out using commercial softwares. The results of the presented study show that the Structure Sensor, which is the low cost structured light 3D sensor, has a great potential for such applications. This study also discusses the benefits and reliability of low-cost structured light systems.

Stochastic 3D modelling of discrete sediment bodies for geotechnical applications

In the course of engineering geological ground characterizations, internal structures of sediments are often simplified to continuous units and structures like lenses are drawn in cross sections based on geologists’ experience. Seeing this as oversimplifications, we propose discrete sediment body (DSB) modelling which is a computational workflow that permits generation of 3D meshes of sediment bodies. DSB modelling can be used to stochastically model highly detailed sedimentary underground structures, based on geometrical assumptions (e.g. size and orientation of lenses) that can be derived from engineering geological investigations. From a geological point of view, the framework is suited to model underground conditions where one would describe the occurrence of chaotic lens-shaped sediments. The modelling process follows a pipeline, where first estimations about the distribution and geometries of the sediment bodies are made. To get non-intersecting meshes as a final product, surrogate point clouds are generated that follow the initial geometrical estimations, which are then clustered, and finally alpha shape based meshes are generated around the clusters. Resulting 3D sediment bodies enable easy generation of consistent 2D geological cross sections and can be used as a basis for further analyses. The paper is finished with a case study where a finite element stability analysis is done for a slope excavation with underground conditions featuring clayey silt lenses in a matrix of gravelly sand. It can be shown that the lenses substantially influence the ground behavior in comparison to a classical approach with continuous units.

ENHANCING GEOMETRIC EDGE DETAILS IN MVS RECONSTRUCTION

Abstract. Mesh models generated by multi view stereo (MVS) algorithms often fail to represent in an adequate manner the sharp, natural edge details of the scene. The harsh depth discontinuities of edge regions are eventually a challenging task for dense reconstruction, while vertex displacement during mesh refinement frequently leads to smoothed edges that do not coincide with the fine details of the scene. Meanwhile, 3D edges have been used for scene representation, particularly man-made built environments, which are dominated by regular planar and linear structures. Indeed, 3D edge detection and matching are commonly exploited either to constrain camera pose estimation, or to generate an abstract representation of the most salient parts of the scene, and even to support mesh reconstruction. In this work, we attempt to jointly use 3D edge extraction and MVS mesh generation to promote edge detail preservation in the final result. Salient 3D edges of the scene are reconstructed with state-of-the-art algorithms and integrated in the dense point cloud to be further used in order to support the mesh triangulation step. Experimental results on benchmark dataset sequences using metric and appearance-based measures are performed in order to evaluate our hypothesis.

3D CONTENT GENERATION USING HYBRID AERIAL SENSOR DATA

Abstract. In aerial data acquisition a new era started with the introduction of the first real hybrid sensor systems, like the Leica CityMapper-2. Hybrid in this context means the combination of an (oblique) camera system with a topographic LiDAR into an integrated aerial mapping system. By combining these complimentary sub-systems into one system the weaknesses of the one system could be compensated by using the alternative data source. An example is the mapping of low-light urban canyons, where image-based systems mostly produce unreliable results. For an LiDAR sensor the geometrical reconstruction of these areas is straight forward and leads to accurate results. The paper gives a detailed overview over the development and technical characteristics of hybrid sensor systems. The process of data acquisition is discussed and strategies for hybrid urban mapping are proposed. A hybrid sensor alone is just a part of the whole procedure to generate 3D content. As important as the senor itself is the workflow to generate the products. Here again a hybrid approach, with the processing of all datasets in one environment, is discussed. Special attention is paid to the hybrid orientation of the data and the integrated generation of base and enhanced products. The paper is rounded off by the discussion of the advantage of LiDAR data for the 3D Mesh generation for urban modelling.

3D hybrid mesh generation with an improved vertical stretch algorithm for geometric models with pinch-out features

Aiming at the infeasibility of using quadrilateral-prisms to describe sharp shapes, an improved vertical stretch algorithm was proposed to generate three-dimensional (3D) hybrid meshes based on the quadrilateral mesh converted from triangles to discretize the geometric models with pinch-out features. A robust and automatic mesh generator was developed, which served as a preprocessing tool for simulating metal forming, water flows and transports. An initial mixed mesh composed of cuspate hexahedra, quadrilateral- and triangular-prisms was constructed to capture the sharp features of pinch-out regions effectively. Five splitting templates were established to divide the 3D elements derived from pinch-out sub-domains into tetrahedra and triangular-prisms. The concept of splitting points was defined, based on which the appropriate implementation mode of each splitting template could be determined, so as to achieve the conformity and consistency of common faces and edges between different kinds of elements. Two topological optimization templates were established to improve the quality of degenerated triangular-prisms and tetrahedra. A new pinch-out boundary quadrilateral insertion method was proposed to avoid the generation of cuspate hexahedra, which simplified the conformity treatment strategies and improved the quality of the resulting mesh to a certain extent. Practical applications confirmed that the hybrid mesh generation methods proposed in this paper could discretize the geometric models with pinch-out features precisely. The requirements on mesh conformity and refinement were also able to be fully satisfied.

Retinal image registration as a tool for supporting clinical applications

Background and Objective: The study of small vessels allows for the analysis and diagnosis of diseases with strong vasculopathy. This type of vessels can be observed non-invasively in the retina via fundoscopy. The analysis of these vessels can be facilitated by applications built upon Retinal Image Registration (RIR), such as mosaicing, Super Resolution (SR) or eye shape estimation. RIR is challenging due to possible changes in the retina across time, the utilization of diverse acquisition devices with varying properties, or the curved shape of the retina.Methods: We employ the Retinal Image Registration through Eye Modelling and Pose Estimation (REMPE) framework, which simultaneously estimates the cameras’ relative poses, as well as eye shape and orientation to develop RIR applications and to study their effectiveness.Results: We assess quantitatively the suitability of the REMPE framework towards achieving SR and eye shape estimation. Additionally, we provide indicative results demonstrating qualitatively its usefulness in the context of longitudinal studies, mosaicing, and multiple image registration. Besides the improvement over registration accuracy, demonstrated via registration applications, the most important novelty presented in this work is the eye shape estimation and the generation of 3D point meshes. This has the potential for allowing clinicians to perform measurements on 3D representations of the eye, instead of doing so in 2D images that contain distortions induced because of the projection on the image space.Conclusions: RIR is very effective in supporting applications such as SR, eye shape estimation, longitudinal studies, mosaicing and multiple image registration. Its improved registration accuracy compared to the state of the art translates directly in improved performance when supporting the aforementioned applications.

Automatic mesh generator for the non-linear homogenized analysis of double curvature masonry structures

An automatic full 3D mesh generator for the non-linear 3D homogenized discrete FE analysis of masonry double curvature structures is presented. The method can reproduce the main features exhibited by masonry, preserving the orthotropy and accounting for the influence of membrane loads on the flexural behavior. Previous homogenization procedures allow reducing the number of variables, leading to non-linear analyses with a low computational burden. In the approach proposed, masonry is modeled through an assemblage of elastic cells joined by non-linear interfaces with finite thickness. The latter are modeled as brick elements with softening replicated by a damage constitutive relationship (known in Abaqus - the commercial FE code used - as Concrete Damage Plasticity). The discrete mesh is obtained automatically by means of an ad-hoc routine which transforms a standard 2D homogeneous FE mesh into a 3D orthotropic elastic model with non-linear interfaces connecting contiguous elastic elements (cells). The proposed method is applied to a hemispherical dome and a cloister vault, with abundant experimental and numerical data available in the literature to compare with. Sensitivity analyses are conducted in order to have an insight into the reliability of the method, which proved excellent predictivity capabilities, as far as both the peak load and the activated failure mechanism are concerned.