3D Shape Research Articles

Deep learning-based prediction of 3-dimensional silver contact shapes enabling improved quality control in solar cell metallization

The industrial metallization of Si solar cells predominantly relies on screen printing, with silver as the preferred electrode material. However, the design of commercial screens often leads to suboptimal silver usage and increased electrical resistance due to print-related inhomogeneities like mesh marks, constrictions and spreading. Real-time monitoring of quality parameters during production has thus become increasingly critical. Current inline optical quality control systems usually only include 2D visualizations of the printed layout, which limits their effectiveness in quality control. Options that allow 3D measurements are usually slow, expensive, and therefore not worth considering in most cases. This research focuses on the development of a model that can estimate the three-dimensional shape of printed contact fingers from a single 2D image without the need of additional hardware using deep learning. Furthermore, a workflow for the generation of training data, which involves the creation of image pairs from a 2D microscope and a 3D confocal laser scanning microscope (CLSM) to accurately represent solar cell fingers, is presented. After model training, the predicted height maps are compared with the ground truth height maps, and the robustness of the model with respect to a paste variation and screen parameter variation is examined. The results confirm the feasibility and reliability of deep learning-based 3D shape estimation, extending its applicability to new, previously unseen data from screen-printed contact fingers. With a structural similarity index (SSIM) score of 0.76, a strong correlation between the estimated and ground truth height maps is established. In summary, our deep learning-based approach for height map estimation offers an effective and reliable solution for fast inline detection and analysis of the cross-sectional area of the printed contact fingers.

Learning to Rasterize Differentiably

AbstractDifferentiable rasterization changes the standard formulation of primitive rasterization — by enabling gradient flow from a pixel to its underlying triangles — using distribution functions in different stages of rendering, creating a “soft” version of the original rasterizer. However, choosing the optimal softening function that ensures the best performance and convergence to a desired goal requires trial and error. Previous work has analyzed and compared several combinations of softening. In this work, we take it a step further and, instead of making a combinatorial choice of softening operations, parameterize the continuous space of common softening operations. We study meta‐learning tunable softness functions over a set of inverse rendering tasks (2D and 3D shape, pose and occlusion) so it generalizes to new and unseen differentiable rendering tasks with optimal softness.

Capacitive Touch Sensing on General 3D Surfaces

Mutual-capacitive sensing is the most common technology for detecting multi-touch, especially on flat and simple curvature surfaces. Its extension to a more complex shape is still challenging, as a uniform distribution of sensing electrodes is required for consistent touch sensitivity across the surface. To overcome this problem, we propose a method to adapt the sensor layout of common capacitive multi-touch sensors to more complex 3D surfaces, ensuring high-resolution, robust multi-touch detection. The method automatically computes a grid of transmitter and receiver electrodes with as regular distribution as possible over a general 3D shape. It starts with the computation of a proxy geometry by quad meshing used to place the electrodes through the dual-edge graph. It then arranges electrodes on the surface to minimize the number of touch controllers required for capacitive sensing and the number of input/output pins to connect the electrodes with the controllers. We reach these objectives using a new simplification and clustering algorithm for a regular quad-patch layout. The reduced patch layout is used to optimize the routing of all the structures (surface grooves and internal pipes) needed to host all electrodes on the surface and inside the object's volume, considering the geometric constraints of the 3D shape. Finally, we print the 3D object prototype ready to be equipped with the electrodes. We analyze the performance of the proposed quad layout simplification and clustering algorithm using different quad meshing and characterize the signal quality and accuracy of the capacitive touch sensor for different non-planar geometries. The tested prototypes show precise and robust multi-touch detection with good Signal-to-Noise Ratio and spatial accuracy of about 1mm.

3Doodle: Compact Abstraction of Objects with 3D Strokes

While free-hand sketching has long served as an efficient representation to convey characteristics of an object, they are often subjective, deviating significantly from realistic representations. Moreover, sketches are not consistent for arbitrary viewpoints, making it hard to catch 3D shapes. We propose 3Dooole, generating descriptive and view-consistent sketch images given multi-view images of the target object. Our method is based on the idea that a set of 3D strokes can efficiently represent 3D structural information and render view-consistent 2D sketches. We express 2D sketches as a union of view-independent and view-dependent components. 3D cubic Bézier curves indicate view-independent 3D feature lines, while contours of superquadrics express a smooth outline of the volume of varying viewpoints. Our pipeline directly optimizes the parameters of 3D stroke primitives to minimize perceptual losses in a fully differentiable manner. The resulting sparse set of 3D strokes can be rendered as abstract sketches containing essential 3D characteristic shapes of various objects. We demonstrate that 3Doodle can faithfully express concepts of the original images compared with recent sketch generation approaches. 1

Length-extended 3D shape sensor using wavelength/space-division multiplexing grating arrays in a multicore fiber.

Limited by the multiplexing number of fiber Bragg grating (FBG), further improvement in the length of 3D shape sensing based on FBG technology is challenging. In this Letter, a wavelength-division and space-division multiplexing multicore fiber grating method is proposed, which extends the sensing length. Employing the femtosecond-laser point-by-point technology, we inscribed WDM grating arrays in six outer cores of a seven-core fiber, respectively. Three cores were utilized as a segment for shape sensing, and two such segments were offset by a specific length and combined to form a shape sensor. Utilizing an FBG interrogator, the proposed shape sensor achieved 2D and 3D shape sensing at a length of 967 mm and effectively mitigated the effects of temperature variations. In experiments, maximum shape reconstruction errors per unit lengths are 1.89%, 2.72%, and 1.47% for 2D shape, 3D shape, and an arbitrary shape under variable temperature conditions, respectively. The proposed method holds promise for further extending the shape sensing length by utilizing multicore fibers or fiber clusters containing more cores.

3D printing assisted surface patterning process on acrylated hydrogels for contact guidance of fibroblasts

Generating stable and customizable topography on hydrogel surfaces with contact guidance potential is critical as it can direct/influence cell growth. This necessitates the development of new techniques for surface patterning of the hydrogels. We report on the design of a square grid template for surface patterning hydrogels. The template was 3-D printed and has the diameter of a well in a 24-well plate. Hyaluronic acid methacrylate (HA) hydrogel precursor solutions were cast on the 3D printed template's surface, which generated 3D square shape topographies on the HA hydrogel surface upon demolding. The 3D Laser Microscopy has shown the formation of a periodic array of 3D topographies on hydrogel surfaces. 3D Laser and Electron Microscopy Imaging have revealed that this new method has increased the surface area and exposed the underlying pore structure of the HA hydrogels. To demonstrate the method's versatility, we have successfully applied this technique to generate 3D topography on two more acrylate hydrogel formulations, gelatin Methacrylate and polyethylene glycol dimethacrylate. Human neonatal dermal fibroblast cells were used as a model cell line to evaluate the cell guidance potential of patterned HA hydrogel. Confocal fluorescence microscopy imaging has revealed that the 3D surface topographies on HA hydrogels can guide and align the actin filaments of the fibroblasts presumably due to the contact guidance mechanism. The newly developed methodology of 3D topography generation in acrylate hydrogels may influence the cell responses on hydrogel surfaces which can impact biomedical applications such as tissue engineering, wound healing, and disease modeling.

Solitary waves of M-fractional low-pass nonlinear electrical transmission line model arising in network system

In this study, we analyze the solitary wave behavior of a truncated M-fractional low-pass nonlinear electrical transmission line (NLETLs) model. NLETL models are relevant to computer network systems, particularly for high-speed data transmissions. They influence the behavior of signals traveling through network cables. To investigate the dynamics of solitary waves in the model, we applied the modified Sardar sub-equation and extended the sinh-Gordon equation expansion methods. We illustrated the 2D, 3D, and contour shapes of selected solutions for appropriate values of the NLETLs dynamics using Mathematica-14. Kink, anti-kink, bright-dark bell, dark bell, M-shaped periodic soliton, and logarithmic wave solutions were obtained. The results indicate that the proposed techniques may provide valuable, powerful, and efficient insights into the dynamics of nonlinear evolution models. The role of the fractional order derivative in making optical solutions is investigated in detail, which opens up opportunities for the creation of more complex models that can more accurately simulate optical phenomena in the real world.

The Role of Different Clay Types in Achieving Low-Carbon 3D Printed Concretes

Concrete 3D printing, an innovative construction technology, offers reduced material waste, increased construction speed, and the ability to create complex and customized shapes that are challenging to achieve with traditional methods. This study delves into the unique fresh-state performance required for 3D printing concrete, discussing buildability, extrudability, and shape retention in terms of concrete rheology, which can be modified using admixtures. Currently most 3D printing concretes feature high cement contents, with little use of secondary cementitious materials. This leads to high embodied carbon, and addressing this is a fundamental objective of this work. The research identifies attapulgite, bentonite, and sepiolite clay as potential concrete admixtures to tailor concrete rheology. Eight low-carbon concrete mixes are designed to incorporate GGBS at a 50% replacement level and are used to measure the influence of each clay on the concrete rheology at varying dosages. A comprehensive rheological test protocol is designed and carried out on all mixes, together with other tests including slump-flow and compression strength. The objective is to determine the applicability of each clay in improving the printability of low-carbon concrete. The findings reveal that at a dosage of 0.5%, sepiolite was seen to improve static yield stress, dynamic yield stress, and rate of re-flocculation, resulting in improved printability. The addition of attapulgite and sepiolite at a dosage of 0.5% by mass of binder increased compressive strength significantly; bentonite had very little influence. These gains are not repeated at 1% clay content, indicating that there may be an optimum clay content. The results are considered encouraging and show the potential of these clays to enhance the performance of low-carbon concrete in 3D printing applications.

FUSION3D: Multimodal data fusion for 3D shape reconstruction: A soft-sensing approach

Traditional high-fidelity imaging techniques, such as X-ray computer tomography (CT), excel in capturing intricate shape details through high-resolution two-dimensional (2D) images. However, the extensive cost or impracticality of obtaining X-ray images poses a significant challenge in various applications (e.g., dense metal components in manufacturing). To overcome this limitation, we propose a soft sensing approach, named FUSION3D, aimed at reconstructing three-dimensional (3D) shapes from 2D sliced images and heterogenous process inputs using a novel model based on multi-modal process knowledge. The developed FUSION3D method is built upon a deep learning framework that utilizes continuous normalizing flow to model the complex shape distribution of 3D objects. We propose a principled probabilistic framework to regularize 3D point clouds by modeling them as a distribution of distributions. Specifically, we learn a two-level hierarchy of distributions where the first level is the distribution of shapes, and the second level is the distribution of points given a shape. Our generative regularization model learns each level of the distribution with a continuous normalizing flow. The invertibility of normalizing flows enables the computation of the likelihood during training and allows us to train our model in the variational inference framework. Our methodology is validated through a case study in additive manufacturing, demonstrating its effectiveness as a soft sensing approach for accurate 3D shape reconstruction without relying on high-fidelity imaging techniques like CT during model deployment.

Strain in Hybrid Organic-Inorganic Metal Halide Perovskites Microstructures by Numerical Simulations.

Hybrid organic-inorganic metal halide perovskites (HOIPs) are promising materials for optoelectronics applications. Their optical and electrical properties can be controlled by strain engineering, that results from application of local elastic deformation or deposition on pre-patterned substrates acquiring a conformal 3D shape. Most interesting, their mechanical properties depend on their crystal structure, composition and dimensionality. We explore by numerical simulations the deformation of a selection of HOIPs comprising a broad range of elastic properties. We consider an axial symmetry with the formation of microdomes on flakes. Radial and vertical forces are considered, finding that the radial force is more effective to obtain large deformation. Large vertical displacement and strain is obtained for HOIPs with low stiffness. The layered nature of HOIPs, that are formed by inorganic layers of different thickness and organic spacers, is also investigated, revealing a non-monotonous trend with the proportion of inorganic to organic part.

SHOWMe: Robust object-agnostic hand-object 3D reconstruction from RGB video

In this paper, we tackle the problem of detailed hand-object 3D reconstruction from monocular video with unknown objects, for applications where the required accuracy and level of detail is important, e.g. object hand-over in human–robot collaboration, or manipulation and contact point analysis. While the recent literature on this topic is promising, the accuracy and generalization abilities of existing methods are still lacking. This is due to several limitations, such as the assumption of known object class or model for a small number of instances, or over-reliance on off-the-shelf keypoint and structure-from-motion methods for object-relative viewpoint estimation, prone to complete failure with previously unobserved, poorly textured objects or hand-object occlusions. To address previous method shortcomings, we present a 2-stage pipeline superseding state-of-the-art (SotA) performance on several metrics. First, we robustly retrieve viewpoints relying on a learned pairwise camera pose estimator trainable with a low data regime, followed by a globalized Shonan pose averaging. Second, we simultaneously estimate detailed 3D hand-object shapes and refine camera poses using a differential renderer-based optimizer. To better assess the out-of-distribution abilities of existing methods, and to showcase our methodological contributions, we introduce the new SHOWMe benchmark dataset with 96 sequences annotated with poses, millimetric textured 3D shape scans, and parametric hand models, introducing new object and hand diversity. Remarkably, we show that our method is able to reconstruct 100% of these sequences as opposed to SotA Structure-from-Motion (SfM) or hand-keypoint-based pipelines, and obtains reconstructions of equivalent or better precision when existing methods do succeed in providing a result. We hope these contributions lead to further research under harder input assumptions. The dataset can be downloaded at https://download.europe.naverlabs.com/showme.

D2GL: Dual-level dual-scale graph learning for sketch-based 3D shape retrieval

Sketch-based 3D shape retrieval (SBSR) is an active research area in the computer vision community, but it is still very challenging. One main reason is that existing deep learning-based methods usually treat sketches as 2D images, neglecting the sparsity and diversity. In this paper, we propose a novel Dual-level Dual-scale Graph Learning (D2GL) method to effectively enhance structural information and produce robust representations for sparse and diverse hand-drawn sketches. Specifically, in addition to the traditional branches for SBSR, we introduce a Dual-level Dual-scale Graph Self-attention (DLDS-GSA) as an auxiliary branch. DLDS-GSA further consists of two levels of encoders, i.e., a local structural encoder and a dual-scale global structural encoder, to capture both local discriminative and multi-scale global structures while minimizing the impact of various sketch drawing details. Comprehensive experiments on SHREC’13 and SHREC’14 datasets demonstrate the superiority of D2GL for SBSR, with extended experiments on PART-SHREC’14 confirming its generalization for unseen classes in SBSR.

Handheld structured light system for panoramic 3D measurement in mesoscale

Abstract The measurement of complete 3D topography in mesoscale plays a vital role in high-precision reverse engineering, oral medical modeling, circuit detection, etc. Traditional structured light systems are limited to measuring 3D shapes from a single perspective. How to achieve high-quality mesoscopic panoramic 3D measurement remains challenging, especially in complex measured scenarios such as dynamic measurement, scattering medium, and high reflectance. To overcome these problems, we develop a handheld mesoscopic panoramic 3D measurement system for such complex scenes together with the fast point-cloud-registration and accurate 3D-reconstruction, where a motion discrimination mechanism is designed to ensure that the captured fringe is in a quasi-stationary case by avoiding the motion errors caused during fringe scanning; a deep neural network is utilized to suppressing the fringe-degradation caused by scattering mediums resulting to significantly improves the quality of the 3D point cloud; a strategy based on phase averaging is additionally proposed to simultaneously correct the saturation-induced errors and gamma nonlinear errors. Finally, the proposed system with a multi-threaded data processing framework is further developed to verify the proposed method and the corresponding experiments verify its feasibility.

Learning intrinsic shape representations via spectral mesh convolutions

We introduce spectral-based convolutional operators embedded within Generalized Graph Neural Networks (G-GNNs). These operators enable deep learning on graphs through a learnable, energy-driven evolution process. This approach empowers us to impose specific properties on the graph convolutional kernel directly derived from the corresponding variational formulations. Our model incorporates both parameterized and non-parameterized graph Laplacian-based energies within the generalized graph convolutional layer to address features like smoothness, sharpness, and compact support. By making appropriate choices within our G-GNN framework, we pave the way for designing novel paradigms for 3D shape representation, reconstruction, and processing, while also enabling effective feature embeddings for intrinsic neural fields.

Flexible circuit-based spatially aware modular optical brain imaging system for high-density measurements in natural settings.

Functional near-infrared spectroscopy (fNIRS) presents an opportunity to study human brains in everyday activities and environments. However, achieving robust measurements under such dynamic conditions remains a significant challenge. The modular optical brain imaging (MOBI) system is designed to enhance optode-to-scalp coupling and provide a real-time probe three-dimensional (3D) shape estimation to improve the use of fNIRS in everyday conditions. The MOBI system utilizes a bendable and lightweight modular circuit-board design to enhance probe conformity to head surfaces and comfort for long-term wearability. Combined with automatic module connection recognition, the built-in orientation sensors on each module can be used to estimate optode 3D positions in real time to enable advanced tomographic data analysis and motion tracking. Optical characterization of the MOBI detector reports a noise equivalence power of 8.9 and at 735 and 850nm, respectively, with a dynamic range of 88dB. The 3D optode shape acquisition yields an average error of 4.2mm across 25 optodes in a phantom test compared with positions acquired from a digitizer. Results for initial in vivo validations, including a cuff occlusion and a finger-tapping test, are also provided. To the best of our knowledge, the MOBI system is the first modular fNIRS system featuring fully flexible circuit boards. The self-organizing module sensor network and automatic 3D optode position acquisition, combined with lightweight modules ( ) and ergonomic designs, would greatly aid emerging explorations of brain function in naturalistic settings.

Cranial Remolding Orthosis Study on the Use of a Temperature Sensor to Measure Wear Time

ABSTRACT Introduction The effectiveness of cranial remolding orthoses (CROs) in treating nonsynostotic deformational plagiocephaly (DP) is anecdotally related to adherence. Efficacy of CROs has been evaluated but lacks evidence regarding necessary wear time to achieve a positive outcome. A 23 hr/d wear schedule is generally prescribed regardless of presentation. The study compares daily wear time to treatment outcomes. Materials and Methods Subjects Infants aged 3–18 months diagnosed with DP were included with initial cranial vault asymmetry (CVA) > 6 mm or cephalic ratio (CR) > 0.90. Subjects were treated with an Orthomerica STARband CRO. A total of 106 subjects enrolled, with 69 completed. Apparatus A questionnaire assessed caregiver’s reported adherence with the CRO. Maximum Integrated’s iButton temperature loggers (iButtons) recorded objective wear time. Procedures 3D head shape measurements via Orthomerica STARScanner, caregivers’ questionnaires, and iButton data were collected every 5–8 weeks. Data Analysis Descriptive statistics were obtained, and data reported as mean ± SD or median [25th, 75th percentiles]. Nonparametric sign tests were used to assess differences from CRO fabrication scan to subsequent follow-ups. Spearman rank correlations and corresponding 95% confidence intervals and P values between average wear time and change of measurements were obtained. Results Self-reported wear time was 22 (22, 23) hrs/d. Measured wear time was 17–18 (12, 20) hrs/d. Longer average wear time was significantly associated with larger reductions in CVA (P = 0.0054), CR (P = 0.0080), and CVAI (P = 0.0059). Conclusions Results showed longer average daily wear increased effectiveness in CRO treatment of plagiocephaly. Increased sample sizing is required to determine if generalizable to brachycephalic and asymmetrical brachycephalic head shapes. Clinical Relevance A CRO wear schedule of 23 hrs a day is currently recommended, but difficult for families to adhere. The impact of the study may indicate a more realistic wear schedule that would improve adherence while achieving optimal outcomes.

Perceptual Biases in the Interpretation of Non-Rigid Shape Transformations from Motion.

Most existing research on the perception of 3D shape from motion has focused on rigidly moving objects. However, many natural objects deform non-rigidly, leading to image motion with no rigid interpretation. We investigated potential biases underlying the perception of non-rigid shape interpretations from motion. We presented observers with stimuli that were consistent with two qualitatively different interpretations. Observers were shown a two-part 3D object with the smaller part changing in length dynamically as the whole object rotated back and forth. In two experiments, we studied the misperception (i.e., perceptual reinterpretation) of the non-rigid length change to a part. In Experiment 1, observers misperceived this length change as a part orientation change (i.e., the smaller part was seen as articulating with respect to the larger part). In Experiment 2, the stimuli were similar, except the silhouette of the part was visible in the image. Here, the non-rigid length change was reinterpreted as a rigidly attached part with an "illusory" non-orthogonal horizontal angle relative to the larger part. We developed a model that incorporated this perceptual reinterpretation and could predict observer data. We propose that the visual system may be biased towards part-wise rigid interpretations of non-rigid motion, likely due to the ecological significance of movements of humans and other animals, which are generally constrained to move approximately part-wise rigidly. That is, not all non-rigid deformations are created equal: the visual systems' prior expectations may bias the system to interpret motion in terms of biologically plausible shape transformations.

Fine-tuning 3D foundation models for geometric object retrieval

Foundation models, such as ULIP-2 (Xue et al., 2023) recently projected forward the field of 3D deep learning. These models are trained with significantly more data and show superior representation learning capacity in many downstream tasks like 3D shape classification and few-shot part segmentation.A particular characteristic of the recent 3D foundation models is that they are typically multi-modal, and involve image (2D) as well as caption (text) branches. This leads to an intricate interplay that benefits all modalities. At the same time, the nature of the 3D encoders alone, involved in these foundation models is not well-understood. Specifically, there is little analysis on the utility of both pre-trained 3D features provided by these models, or their capacity to adapt to new downstream 3D data. Furthermore, existing studies typically focus on label-oriented downstream tasks, such as shape classification, and ignore other critical applications, such as 3D content-based object retrieval.In this paper, we fill this gap and show, for the first time, how 3D foundation models can be leveraged for strong 3D-to-3D retrieval performance on seven different datasets, on par with state-of-the-art view-based architectures. We evaluate both the pre-trained foundation models, as well as their fine-tuned versions using downstream data. We compare supervised fine-tuning using classification labels against two self-supervised label-free fine-tuning methods. Importantly, we introduce and describe a methodology for fine-tuning, as we found this to be crucial to make transfer learning from 3D foundation models work in a stable manner.

Curvy cuts: Programming axisymmetric kirigami shapes

Although bending a sheet of paper is an easy operation, stretching is more limited and it leads to rupture and tears. However, well-designed cuts on the sheet can induce a large effective stretchability. This kirigami technique offers a large scope of engineering applications ranging from deployable structures to compliant electronics. We are here interested in the axisymmetric configuration where cuts are designed along concentric circles. Applying an increasing transverse load at the center of the sheet results into a 3D axisymmetric structure of growing amplitude which eventually saturates. We first describe the linear response of the structure and determine the evolution of the deployed shape until its asymptotic geometrical limit. Reversing the problem in the linear regime, we propose, a design procedure for the cuts leading to a desired 3D shape. The structure can also be deployed by inflating an inner balloon. Exploring further the interplay between mechanics and geometry, we finally describe the maximum volume of inflated kirigami structures as a function of the cutting pattern.

The continuous adjoint method to the γ−R˜eθt$$ \gamma -\tilde{R}{e}_{\theta t} $$ transition model coupled with the Spalart–Allmaras model for compressible flows

AbstractThe continuous adjoint method for transitional flows of compressible fluids is developed and assessed, for the first time in the literature. The gradient of aerodynamic objective functions (aerodynamic forces) with respect to design variables, in problems governed by the compressible Navier–Stokes equations coupled with the Spalart–Allmaras turbulence model and the transition model (in three, non‐smooth and smooth, variants of it), is computed based on the continuous adjoint method. The development of the adjoint to the smooth transition model variant proved to be beneficial. The accuracy of the computed sensitivity derivatives is verified against finite differences. Programming is performed in an in‐house, vertex‐centered finite‐volume code, efficiently running on GPUs. The proposed continuous adjoint method is used in 2D and 3D aerodynamic shape optimization problems, namely the constrained optimization of the NLF(1)–0416 isolated airfoil and that of the ONERA M6 wing. The impact of “frozen transition” (assumption according to which the adjoint to the transition model equations are not solved) or “frozen turbulence” (by additionally ignoring the adjoint to the turbulence model) are evaluated; it is shown that both lead to inaccurate sensitivities.