Shape Classification Research Articles

Comparing Shape Representations for the Aesthetic Judgement of 3D Shape Pairs.

Visual aesthetics of 3D shapes is a fundamental perceptual attribute. In this paper, we explore the question of how different shape representations affect the aesthetic judgments of shape pairs. Specifically, we compare human responses to perceptual aesthetics judgements on 3D shapes in pairs presented in different 3D shape representation such as voxels, points, wireframe, and polygon. In contrast to our own previous work [8] that explores this question for a few shape classes, this paper analyzes a larger number of shape classes. Our key finding is that human aesthetics judgements on relatively low resolution of points or voxels are comparable to polygon meshes, which implies that humans can often make their aesthetics decisions based on relatively coarse representations of the shapes. Our results have implications towards the data collection process of pairwise aesthetics data and the further use of such data in shape aesthetics and 3D modeling problems.

Coronal, Sagittal, and Horizontal Classification of the Chest Shape and Its Role in Selection of Proper Implants.

The number of female breast correction procedures has been steadily increasing. Despite extensive literature being available on these procedures, few authors have discussed the role of preoperative chest wall analysis in relation to postoperative outcomes. To date, no comprehensive classification of chest shape has been introduced in the literature. The aim of this study was to present a novel classification of chest shapes, based on three basic planes: coronal (C), sagittal (S), and horizontal (H), hence the proposed name "CSH classification." In this study, a retrospective analysis of 1000 randomly selected chest and breast images was conducted by three independent nonmedical evaluators, using standardized digital images captured with the Vectra 3D body scanner. All examined patients were qualified for breast augmentation surgery. Among 1000 randomly examined patients, 923 were classified in the coronal plane, 920 in the sagittal plane, and 627 in the horizontal plane. Other patients were excluded from the study due to insufficient quality of the images. A notable 43.2% of women have shoulder height asymmetries. A relationship between shoulder height asymmetry and chest width was confirmed by the chi-square Pearson test (P = 0.04), indicating that a higher shoulder is associated with a greater chest width on the same side. Furthermore, 84.7% of women displayed excessive upper chest projection, whereas 28.4% showed excessive lower chest projection. Additionally, 84.4% of women had chest projection asymmetries. The vast majority of the women had natural asymmetries of the breast or and chest wall. The CSH classification allows systematizing the chest shape assessment. The chest shape has a significant impact on breast implant selection and the choice of the breast surgery technique.

A Deeper Analysis of Volumetric Relightable Faces.

Portrait viewpoint and illumination editing is an important problem with several applications in VR/AR, movies, and photography. Comprehensive knowledge of geometry and illumination is critical for obtaining photorealistic results. Current methods are unable to explicitly model in 3D while handling both viewpoint and illumination editing from a single image. In this paper, we propose VoRF, a novel approach that can take even a single portrait image as input and relight human heads under novel illuminations that can be viewed from arbitrary viewpoints. VoRF represents a human head as a continuous volumetric field and learns a prior model of human heads using a coordinate-based MLP with individual latent spaces for identity and illumination. The prior model is learned in an auto-decoder manner over a diverse class of head shapes and appearances, allowing VoRF to generalize to novel test identities from a single input image. Additionally, VoRF has a reflectance MLP that uses the intermediate features of the prior model for rendering One-Light-at-A-Time (OLAT) images under novel views. We synthesize novel illuminations by combining these OLAT images with target environment maps. Qualitative and quantitative evaluations demonstrate the effectiveness of VoRF for relighting and novel view synthesis, even when applied to unseen subjects under uncontrolled illumination. This work is an extension of Rao et al. (VoRF: Volumetric Relightable Faces 2022). We provide extensive evaluation and ablative studies of our model and also provide an application, where any face can be relighted using textual input.

DTV-CNN: Neural network based on depth and thickness views for efficient 3D shape classification

Fast and effective algorithms for deep learning on 3D shapes are keys to innovate mechanical and electronic engineering design workflow. In this paper, an efficient 3D shape to 2D images projection algorithm and a shallow 2.5D convolutional neural network architecture is proposed. A smaller convolutional neural network (CNN) model is achieved by information enrichment at the preprocessing stage, i.e. 3D geometry is compressed into 2D “thickness view” and “depth view”. Fusing the depth view and thickness view (DTV) from the same projection view into a dual-channel grayscale image, can improve information locality for geometry and topology feature extraction. This approach bridges the gap between mature image deep learning technologies to the applications of 3D shape. Enhanced by several essential scalar geometry properties and only 3 projection views, a mixed CNN and multiple linear parameter (MLP) neural network model achives a validation accuracy of 92 % for ModelNet10 mesh-based dataset, while the training time is one order of magnitude less than the original multi-view CNN approach. This study also creates new 3D shape datasets from 2 open source CAD projects. Higher validation accuracy is obtained for realistic CAD datasets, i.e. 97 % for FreeCAD's mechanical part library and 95 % for KiCAD electronic part library. The training cost reduces to tens of minutes on a laptop CPU, given the smaller input data size and shallow neural network design. It is expected that this approach can be adapted for other machine learning scenarios involved in CAD geometry.

Nanoparticles with cubic symmetry: classification of polyhedral shapes

Structural studies of polyhedral bodies can help to analyze geometric details of observed crystalline nanoparticles (NP) where we consider compact polyhedra of cubic point symmetry as simple models. Their surfaces are described by facets with normal vectors along selected Cartesian directions (a, b, c) together with their symmetry equivalents forming a direction family {abc}. Here we focus on polyhedra with facets of families {100}, {110}, and {111}, suggested for metal and oxide NPs with cubic lattices. Resulting generic polyhedra, cubic, rhombohedral, octahedral, and tetrahexahedral, have been observed as NP shapes by electron microscopy. They can serve for a complete description of non-generic polyhedra as intersections of corresponding generic species, not studied by experiment so far. Their structural properties are shown to be fully determined by only three parameters, facet distances R 100, R 110, and R 111 of the three facet types. This provides a novel phase diagram to systematically classify all corresponding polyhedra. Their structural properties, such as shape, size, and facet geometry, are discussed in analytical and numerical detail with visualization of typical examples. The results may be used for respective NP simulations but also as a repository stimulating the structural interpretation of new NP shapes to be observed by experiment.

On a numerical approach for solving some geometrical shape optimization problems in fluid mechanics

This paper is devoted to the numerical study of geometrical shape optimization problems in fluid mechanics, which consist in minimizing some criterion volume cost functionals on a family of admissible doubly connected domains, constrained by steady-state Stokes boundary value problems. We establish the existence of the shape derivative of the considered cost functionals, by means of Minkowski deformation, using the shape derivative formulas, recently established in Boulkhemair and Chakib (2014). This allows us to express the shape derivative by means of the support function and to avoid the tedious computations required when one use the gradient optimization process based on the classical shape derivative, involving the vector fields, notably when one opt for the finite element discretization. So, based on the established shape derivative formulas, we propose a shape optimization numerical process for solving these problems, using the gradient descent algorithm performed by the finite element discretization, for approximating the auxiliary boundary value Stokes problems. Finally, in order to show the validity and the effectiveness of the proposed approach, we present some numerical tests obtained by solving some shape optimization problems of minimizing different cost functionals on various configurations of domains, constrained by steady-state stokes boundary value problems with different boundary conditions. These numerical simulations include some comparison results showing that the proposed approach is more efficient than the gradient approach based on the classical shape derivative, in terms of the accuracy of the solution and central processing unit (CPU) time execution.

STED microscopy reveals dendrite-specificity of spines in turtle cortex

Dendritic spines are key structures for neural communication, learning and memory. Spine size and shape probably reflect synaptic strength and learning. Imaging with superresolution STED microscopy the detailed shape of the majority of the spines of individual neurons in turtle cortex (Trachemys scripta elegans) revealed several distinguishable shape classes. Dendritic spines of a given class were not distributed randomly, but rather decorated significantly more often some dendrites than others. The individuality of dendrites was corroborated by significant inter-dendrite differences in other parameters such as spine density and length. In addition, many spines were branched or possessed spinules. These findings may have implications for the role of individual dendrites in this cortex.

Recognition of building shape in maps using deep graph filter neural network

Shape is one of the core features of the buildings which are the main elements of the map. The building shape recognition is widely used in many spatial applications. Due to the irregularity of the building contour, it is still challenging for building shape recognition. Inspired by graph signal processing theory, we propose a deep graph filter neural network (DGFN) for the shape recognition of buildings in maps. First, we regard shape recognition as a combination of subjective and objective graph signal filtering process. Second, we construct a shape features extraction framework from the perspective of shape details, shape structure and shape local information. Third, DGFN model can fulfil the tasks of shape classification and shape embedding of building at the same time. Finally, multi angle experiments verify our viewpoint of shape recognition mechanism, and the comparison with similar algorithms proves the high accuracy and availability of DGFN model.

Can Distal First Metatarsal Supination Osteotomy, Varization Osteotomy, or a Combination of Both Improve HV Surgical Outcomes? Preliminary Results of a Multicenter Randomized Control Trial

Category: Bunion; Midfoot/Forefoot Introduction/Purpose: The outcomes of Hallux Valgus (HV) surgical treatment are perfectible. First ray dysmorphia in HV appears to incorporate a first metatarsal (M1) hyperpronation and a valgus deviation of M1 distal articular surface. Those could play a role in first ray destabilization and misalignment in HV. However, we do not know whether correction of these deformities could improve HV surgical treatment outcomes. Therefore, we performed a study to assess the impact of M1 distal supination osteotomy, M1 distal varization osteotomy, and combination of both on HV clinical and radiological outcomes. We hypothesized that M1 distal supination osteotomies would improve surgical outcomes in HV. Methods: We conducted an IRB-approved multicenter randomized control trial. HV with an indication for surgical correction were included. HV requiring surgery of the lesser metatarsals were excluded. Feet were randomly divided into 4 groups corresponding to 4 different types of surgery (Figure): 1. Control, classical M1 distal chevron. 2. Supination M1 distal chevron 3. Varization M1 distal chevron 4. Combination of the latter two. Preoperative evaluation and 6-months follow-up were performed. Demographics, AOFAS score, complications, and patient satisfaction (Likert Scale) were collected. On conventional weight-bearing radiographs before and 6 months after surgery, we measured the Hallux Valgus Angle(HVA), the Intemetatarsal Angle(IMA) the Distal Metatarsal Articular Angle(DMAA), the Okuda's M1 head shape classification(from round to angular) and the sesamoid position according to Hardy and Clapham. Normality of different variables was assessed using the Shapiro-Wilk test. Groups were compared using ANOVA for normal and Kruskall Wallis for nonnormal variables. Post-hoc pairwise analyses were performed with Dunn-Bonferoni’s test. Results: 100 HV were included. 8 were lost to follow-up. The Control (n=24), Supination (n=20), Varization (n=26) and Combination (n=22) groups were comparable on age, gender, BMI, and preoperative radiological parameters. At 6 months, there was no difference on AOFAS improvement (p=0.39) and patients satisfaction (p=0.14) whereas there were significant differences on HVA (p < 0.001), IMA (p < 0.001), DMAA (p < 0.001), Okuda's classifcation (p < 0.001) and sesamoid position (p < 0.001) between groups. We found lower HVA values for Combination VS Control (p < 0.001), Combination VS Varization (p < 0.001), Supination VS Control (p < 0.001) and Supination VS Varization (p = 0.005). Asymptomatic recurrences (HVA>15°) were present in 5 cases in Control (20.8%), and 2 in Varization (7.7%). Asymptomatic Hallux Varus (HVA < 0°) were present in 4 cases in Combination (18.2%), and 1 in Supination (5%). Conclusion: Distal M1 supination osteotomies showed better improvement in radiological parameters without showing clinical benefit. The combination of supination and varization osteotomies could lead to overcorrection while the classic M1 distal chevron could lead to insufficient corrections. Routinely performing a supination osteotomy in HV surgical management may lead to erroneous correction and we recommend an à la carte treatment. These results are preliminary and we recommend more patients and longer follow-up to confirm these findings.

Anatomically Guided Cross-Domain Repair and Screening for Ultrasound Fetal Biometry.

Ultrasound based estimation of fetal biometry is extensively used to diagnose prenatal abnormalities and to monitor fetal growth, for which accurate segmentation of the fetal anatomy is a crucial prerequisite. Although deep neural network-based models have achieved encouraging results on this task, inevitable distribution shifts in ultrasound images can still result in severe performance drop in real world deployment scenarios. In this article, we propose a complete ultrasound fetal examination system to deal with this troublesome problem by repairing and screening the anatomically implausible results. Our system consists of three main components: A routine segmentation network, a fetal anatomical key points guided repair network, and a shape-coding based selective screener. Guided by the anatomical key points, our repair network has stronger cross-domain repair capabilities, which can substantially improve the outputs of the segmentation network. By quantifying the distance between an arbitrary segmentation mask to its corresponding anatomical shape class, the proposed shape-coding based selective screener can then effectively reject the entire implausible results that cannot be fully repaired. Extensive experiments demonstrate that our proposed framework has strong anatomical guarantee and outperforms other methods in three different cross-domain scenarios.

Image classification of Human Face Shapes Using Convolutional Neural Network Xception Architecture with Transfer Learning

Abstract. The development of information technology in facial recognition is influenced by a faster and more accurate authentication system. This allows the computer system to identify a person's face.
 Purpose: Similar to fingerprints and the retina of the human eye, each person's face has a different shape and contour. Since it is known that the human face provides a lot of information, as well as topics that attract attention make it studied intensively.
 Methods/Study design/approach: Several studies examining information from human faces are facial recognition. One of the approaches used to recognize facial imagery is through the use of a Convolutional Neural Network (CNN). CNN is a method in the field of Deep Learning that can be used to recognize and classify objects in digital images. In this study, the method used to implement facial image classification is the Xception architecture CNN algorithm with a transfer learning approach.
 Result/Findings: The dataset used in this study was obtained from Kaggle, namely the Face Shape Dataset which contains 5000 data. After testing, an accuracy rate of 96.2% was obtained in the training process and 81.125% in the validation process. This study also uses new data to test the model that has been made, and the results show an accuracy rate of 85.1% in classifying facial imagery.
 Novelty/Originality/Value: Therefore, it can be said that the model created in this study has the ability to classify images of facial shapes Human Face Shapes Using Convolutional Neural Network Xception Architecture with Transfer Learning.

Assessment of Bladder Filling Type on Radiation Dose to Organs at Risk in MR-Guided Intracavity Brachytherapy for Cervical Cancer

In this study, we evaluated whether the classification of bladder shape affects the absorbed dose (Gy) of OARs and its geometrical position to normal organs in MR-guided intracavitary brachytherapy (ICBT). In this study, 269 patients who underwent MR-guided ICBT for cervical cancer from 2016 to 2022 were included. The applicator-inserted bladder filling type (BFT) classification was divided into three types (tilted, curved, and other shapes: group E). The anatomical positional relationship between the uterus and its surroundings was measured on pre-MR images and ICBT simulation MR images. Spearman's rank correlation analysis was used for bladder volume and OAR dose according to BFT. Anatomical differences were analyzed by ANOVA by measuring the distance from the center to both bladder walls. In the correlation analysis considering the shape of the bladder, the tilted, curved, and group E had Spearman's ρ of 0.211; -0.323, 0.412; -0.307 and -0.035; -0.209 for maximum absorbed dose (D2cc) of bladder and small bowel, respectively. It was statistically significant in the tilted type and curved type. The average left and right bladder lengths differences of the tilted type were the highest at 8.47 cm and 7.11 cm, respectively. It was a statistically significant between bladder shape and the difference in left and right bladder lengths differences (p< 0.01). In this study, it was confirmed that bladder distension increased the maximum bladder dose (D2cc) and decreased the maximum bowel dose (D2cc) to the tilted type and curved type. In addition, if the left and right bladder lengths are measured in pre-MR, the degree of bladder distension can be evaluated in advance by checking the shape of the bladder in the case of the tilting type before ICBT. Based on these findings, a prospective study is needed to evaluate the effect of ICBT on cervical cancer treatment outcomes through bladder type classification.

Research on the Shape Classification Method of Rural Homesteads Based on Parcel Scale—Taking Yangdun Village as an Example

The basic information survey on homesteads requires understanding the shape of homesteads, and the shape of the homesteads based on the spatial location can reflect information such as their outline and regularity, but the current shape classification of rural homesteads at the parcel scale lacks analytical methods. In this study, we endeavor to explore a classification model suitable for characterizing homestead shapes at the parcel scale by assessing the impact of various research methods. Additionally, we aim to uncover the evolutionary patterns in homestead shapes. The study focuses on Yangdun Village, located in Deqing County, Zhejiang Province, as the research area. The data utilized comprise Google Earth satellite imagery and a vector layer representing homesteads at the parcel scale. To classify the shapes of homesteads and compare classification accuracy, we employ a combination of methods, including the fast Fourier transform (FFT), Hu invariant moments (HIM), the Boyce and Clark shape index (BCSI), and the AlexNet model. Our findings reveal the following: (1) The random forest method, when coupled with FFT, demonstrates the highest effectiveness in identifying the shape categories of homesteads, achieving an average accuracy rate of 88.6%. (2) Combining multiple methods does not enhance recognition accuracy; for instance, the accuracy of the FFT + HIM combination was 88.4%. (3) The Boyce and Clark shape index (BCSI) proves unsuitable for classifying homestead shapes, yielding an average accuracy rate of only 58%. Furthermore, there is no precise numerical correlation between the homestead category and the shape index. (4) It is noteworthy that over half of the homesteads in Yangdun Village exhibit rectangular-like shapes. Following the “homesteads reform”, square-like homesteads have experienced significant vacating, resulting in a mixed arrangement of homesteads overall. The research findings can serve as a methodological reference for the investigation of rural homestead shapes. Proficiency in homestead shape classification holds significant importance in the realms of information investigation, regular management, and layout optimization of rural land.

3D shape classification based on global and local features extraction with collaborative learning

3D shape classification based on global and local features extraction with collaborative learning

JWST photometry of globular cluster populations in Abell 2744 at z = 0.3

ABSTRACT JWST imaging of the rich galaxy cluster Abell 2744 at z = 0.308 has been used by the UNCOVER team to construct mosaic images in the NIRCam filters. The exceptionally deep images in the (F115W, F150W, and F200W) bands reveal a large population of unresolved point-like sources across the field, the vast majority of which are globular clusters (GCs) or ultra-compact dwarfs. To the limits of our photometry, more than 10 000 such objects were measured, most of which are in the haloes of the five largest A2744 galaxies but which also include GCs around some satellite galaxies and throughout the intergalactic medium. The measured luminosity function covers almost 4 mag and follows a classic lognormal shape, though the GCLF turnover point still lies fainter than the photometric completeness limit. The colour index (F115W–F200W) in particular covers an intrinsic spread of 0.5 mag, clearly resolving the expected range of GC metallicities. The overall results are consistent with a large, normal GC population seen at a 3.5 Gyr earlier stage of dynamical evolution.

Stability of rotating equilibrium states of fluid deformable surfaces

AbstractWe consider rotating equilibrium states of fluid deformable surfaces. These states are characterized by a force balance between centrifugal and bending forces, involve surface Killing vector fields and are independent of surface viscosity. Considering a continuum description based on the incompressible surface Navier‐Stokes equations with bending forces and conserved enclosed volume we numerically demonstrate that these rotating equilibrium states can be reached, but also that these states are not stable. Any perturbation in shape or rotating flow field leads to dissipation and destroys the rotating equilibrium states. After breaking symmetry, evolution reaches other rotating states with a lower energy for which the symmetry axis and the rotation axis are not aligned. Such flow fields could be characterized by three‐dimensional Killing vector fields. However, also these states are also not stable. Based on these numerical results we postulate a cascading mechanism of ‘disturbance ‐ force balance reconfiguration ‐ dissipation’ that contains various rotating equilibrium states as transient configurations but eventually leads to the classical equilibrium shapes of the Helfrich energy.

A viewpoint-guided prototype network for 3D shape classification

A viewpoint-guided prototype network for 3D shape classification

Distinguishing and Matching-Aware Unsupervised Point Cloud Completion

Real-scanned point clouds are often incomplete due to occlusion, light reflection and limitations of sensor resolution, which impedes the related progress of downstream tasks, e.g., shape classification and object detection. Although there has been impressive research progress on the point cloud completion topic, they rely on the premise of extensive paired training data. However, collecting complete point clouds in some specified scenarios is labor-intensive and even impractical. To mitigate this problem, we propose DMNet, a distinguishing and matching-aware unsupervised point cloud completion network. Our work belongs to the group of unsupervised completion methods but goes beyond previous studies. Firstly, we propose a distinguishing-aware feature extractor to learn discriminable semantic information for different instances, simultaneously enhancing the robust invariant representation under noise disturbances. Secondly, we design a hierarchy-aware hyperbolic decoder to recover the complete geometry of point clouds, which not only can capture the implicit hierarchical relationships in data but also has an explicit extended nature. Finally, we develop a matching-aware refiner to eliminate noise points via aligning the topology structure of the input and predicted partial point clouds. Extensive experiments on MVP, Completion3D and KITTI datasets prove the effectiveness of our method, which performs favorably over state-of-the-art methods both quantitatively and qualitatively.

Coarse Aggregate Shape Classification Method Based on Per-Optuna-LightGBM Model

To improve the detection level of aggregate shape for automated road use, Per-Optuna-LightGBM model for aggregate shape classification is proposed. Collect aggregate images using industrial camera and extract 48 morphological feature parameters. A feature importance analysis method based on Spearman Correlation and Permutation Importance is proposed to remove redundant factors and select the feature parameters of aggregate morphology. Based on cross-validation, an optimized Optuna-LightGBM model is trained based on the constructed dataset. Compared with GS-XGBoost algorithm, the Optuna-LightGBM model can classify the shape of aggregates more accurately and efficiently. The accuracy value of the proposed model is 82.5%, which increased by 4% compared to before optimization. The proposed model can efficiently classify the shape of aggregates which meet the design requirements, also provide a certain foundation for automated classification of aggregate shapes.

LCPFormer: Towards Effective 3D Point Cloud Analysis via Local Context Propagation in Transformers

Transformer with its underlying attention mechanism and the ability to capture long-range dependencies makes it become a natural choice for unordered point cloud data. However, local regions separated from the general sampling architecture corrupt the structural information of the instances, and the inherent relationships between adjacent local regions lack exploration. In other words, the transformer only focuses on the long-range dependence, while local structural information is still crucial in a transformer-based 3D point cloud model. To enable transformers to incorporate local structural information, we proposed a straightforward solution based on the natural structure of the point clouds to exploit the message passing between neighboring local regions, thus making their representations more comprehensive and discriminative. Concretely, the proposed module, named Local Context Propagation (LCP), is inserted between two transformer layers. It takes advantage of the overlapping points of adjacent local regions (statistically shown to be prevalent) as intermediaries, then re-weighs the features of these shared points from different local regions before passing them to the next layers. Finally, we design a flexible LCPFormer architecture equipped with the LCP module, which is applicable to several different tasks. Experimental results demonstrate that our proposed LCPFormer outperforms various transformer-based methods in benchmarks including 3D shape classification and dense prediction tasks such as 3D object detection and semantic segmentation. Code will be released for reproduction.