Shape Classification Research Articles

Quantitative 2D fruit shape analysis of a wide range of pear genetic resources toward shape design breeding

The shape of a fruit is one of the main characteristics of fruit morphology, and its quantitative description is crucial to improve fruit quality. However, the shape of a pear fruit is relatively complex, and difficult to describe comprehensively. This study aimed to quantitatively analyze the morphological diversity of 2D fruit shapes of 276 cultivars, covering 6 groups including Pyrus pyrifolia, Pyrus bretschneideri, Pyrus ussuriensis, Pyrus sinkiangensis, Pyrus communis and interspecific hybridization, and discover shape traits with high heritability. Through an automated image processing method, the fruit contour was first accurately extracted, and then geometrical parameters and Fourier descriptors of the pear fruit shape were obtained. Furthermore, principal components (PCs) of the Fourier descriptors were analyzed, and the broad-sense heritability (H2) of the PCs and standardized width (WSD) were estimated. After that, the pear shape stability was calculated, and cultivars were clustered by their shape characteristics. This paper mainly discusses the longitudinal section, as the cross-section variation among cultivars is small. For the longitudinal section, the results showed more than 95 % of the two-dimensional shape variation can be explained by only six principal components; WSD20, WSD30 and PC1 have high H2; the shape stability of western pear was the lowest; a new class of shape was found by clustering the shape traits. The results indicated that the proposed pipeline can provide an objective and comprehensive quantitative description of pear fruit shape for a wide range of cultivars, which is necessary to discover the key genes determining pear shape. This approach is also beneficial for research on pear breeding and germplasm resource management.

Microplastics in Qinghai-Tibet Plateau and Yunnan-Guizhou Plateau lakes, China

As an emerging environmental pollutant, microplastics (MPs) have received widespread attention. Recently, studies examining microplastic pollution in plateau lakes have been increasing, but few have examined the distributions, sources, and fates of MPs in different plateau areas. In this work, the abundances and characteristics of MPs in surface waters and sediments in lakes of the Qinghai-Tibet Plateau (QTP) and Yunnan-Guizhou Plateau (YGP) were systematically investigated. The abundances of MPs in the lakes of the QTP ranges within 0.05–1.8 n/L in surface waters and 10–2643.7 n/kg in sediments. In the lakes of the YGP, the abundances of MPs ranged within 1.3–10.1 n/L in surface waters and 171.7–4260 n/kg in sediments. The dominant shape, color, and size class of MPs were fiber, transparent, and 0–0.5 mm in plateau lakes, respectively. MPs were mainly composed of polypropylene, polyethylene, and polyethylene terephthalate polymers. The different sources of MPs in the QTP and YGP lakes were mainly due to differences in human activities. The primary sources of microplastic pollution in the lakes of the QTP were tourism and atmospheric transport, while sewage discharge, agriculture, and fishing activities were the main sources of MPs in urban lakes of the YGP. Although the level of microplastic pollution in plateau lakes was relatively low, the sources should be identified and monitored so that the effects and extent of microplastic pollution in these fragile environments can be fully understood. This study provides a valuable dataset and theoretical basis for subsequent research on microplastic pollution in plateau lakes.

Sharp-Interface Limit of a Multi-phase Spectral Shape Optimization Problem for Elastic Structures

We consider an optimization problem for the eigenvalues of a multi-material elastic structure that was previously introduced by Garcke et al. (Adv. Nonlinear Anal. 11:159–197, 2022). There, the elastic structure is represented by a vector-valued phase-field variable, and a corresponding optimality system consisting of a state equation and a gradient inequality was derived. In the present paper, we pass to the sharp-interface limit in this optimality system by the technique of formally matched asymptotics. Therefore, we derive suitable Lagrange multipliers to formulate the gradient inequality as a pointwise equality. Afterwards, we introduce inner and outer expansions, relate them by suitable matching conditions and formally pass to the sharp-interface limit by comparing the leading order terms in the state equation and in the gradient equality. Furthermore, the relation between these formally derived first-order conditions and results of Allaire and Jouve (Comput. Methods Appl. Mech. Eng. 194:3269–3290, 2005) obtained in the framework of classical shape calculus is discussed. Eventually, we provide numerical simulations for a variety of examples. In particular, we illustrate the sharp-interface limit and also consider a joint optimization problem of simultaneous compliance and eigenvalue optimization.

Learning Cross-Attention Point Transformer With Global Porous Sampling.

In this paper, we propose a point-based cross-attention transformer named CrossPoints with parametric Global Porous Sampling (GPS) strategy. The attention module is crucial to capture the correlations between different tokens for transformers. Most existing point-based transformers design multi-scale self-attention operations with down-sampled point clouds by the widely-used Farthest Point Sampling (FPS) strategy. However, FPS only generates sub-clouds with holistic structures, which fails to fully exploit the flexibility of points to generate diversified tokens for the attention module. To address this, we design a cross-attention module with parametric GPS and Complementary GPS (C-GPS) strategies to generate series of diversified tokens through controllable parameters. We show that FPS is a degenerated case of GPS, and the network learns more abundant relational information of the structure and geometry when we perform consecutive cross-attention over the tokens generated by GPS as well as C-GPS sampled points. More specifically, we set evenly-sampled points as queries and design our cross-attention layers with GPS and C-GPS sampled points as keys and values. In order to further improve the diversity of tokens, we design a deformable operation over points to adaptively adjust the points according to the input. Extensive experimental results on both shape classification and indoor scene segmentation tasks indicate promising boosts over the recent point cloud transformers. We also conduct ablation studies to show the effectiveness of our proposed cross-attention module with GPS strategy.

Bridging Actions: Generate 3D Poses and Shapes In-Between Photos.

Generating realistic 3D human motion has been a fundamental goal of the game/animation industry. This work presents a novel transition generation technique that can bridge the actions of people in the foreground by generating 3D poses and shapes in-between photos, allowing 3D animators/novice users to easily create/edit 3D motions. To achieve this, we propose an adaptive motion network (ADAM-Net) that effectively learns human motion from masked action sequences to generate kinematically compliant 3D poses and shapes in-between given temporally-sparse photos. Three core learning designs underpin ADAM-Net. First, we introduce a random masking process that randomly masks images from an action sequence and fills masked regions in latent space by interpolation of unmasked images to simulate various transitions under given temporally-sparse photos. Second, we propose a long-range adaptive motion (L-ADAM) attention module that leverages visual cues observed from human motion to adaptively recalibrate the range that needs attention in a sequence, along with a multi-head cross-attention. Third, we develop a short-range adaptive motion (S-ADAM) attention module that weightedly selects and integrates adjacent feature representations at different levels to strengthen temporal correlation. By coupling these designs, the results demonstrate that ADAM-Net excels not only in generating 3D poses and shapes in-between photos, but also in classic 3D human pose and shape estimation.

PointMCD: Boosting Deep Point Cloud Encoders Via Multi-View Cross-Modal Distillation for 3D Shape Recognition

As two fundamental representation modalities of 3D objects, 3D point clouds and multi-view 2D images record shape information from different domains of geometric structures and visual appearances. In the current deep learning era, remarkable progress in processing such two data modalities has been achieved through respectively customizing compatible 3D and 2D network architectures. However, unlike multi-view image-based 2D visual modeling paradigms, which have shown leading performance in several common 3D shape recognition benchmarks, point cloud-based 3D geometric modeling paradigms are still highly limited by insufficient learning capacity due to the difficulty of extracting discriminative features from irregular geometric signals. In this article, we explore the possibility of boosting deep 3D point cloud encoders by transferring visual knowledge extracted from deep 2D image encoders under a standard teacher-student distillation workflow. Generally, we propose PointMCD, a unified multi-view cross-modal distillation architecture, including a pretrained deep image encoder as the teacher and a deep point encoder as the student. To perform heterogeneous feature alignment between 2D visual and 3D geometric domains, we further investigate visibility-aware feature projection (VAFP), by which point-wise embeddings are reasonably aggregated into view-specific geometric descriptors. By pair-wisely aligning multi-view visual and geometric descriptors, we can obtain more powerful deep point encoders without exhausting and complicated network modification. Experiments on 3D shape classification, part segmentation, and unsupervised learning strongly validate the effectiveness of our method. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">The code and data will be publicly available at</i> <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/keeganhk/PointMCD</uri> .

Point Cloud Classification Using Content-Based Transformer via Clustering in Feature Space

Recently, there have been some attempts of Transformer in 3D point cloud classification. In order to reduce computations, most existing methods focus on local spatial attention, but ignore their content and fail to establish relationships between distant but relevant points. To overcome the limitation of local spatial attention, we propose a point content-based Transformer architecture, called PointConT for short. It exploits the locality of points in the feature space (content-based), which clusters the sampled points with similar features into the same class and computes the self-attention within each class, thus enabling an effective trade-off between capturing long-range dependencies and computational complexity. We further introduce an Inception feature aggregator for point cloud classification, which uses parallel structures to aggregate high-frequency and low-frequency information in each branch separately. Extensive experiments show that our PointConT model achieves a remarkable performance on point cloud shape classification. Especially, our method exhibits 90.3% Top-1 accuracy on the hardest setting of ScanObjectNN. Source code of this paper is available at https://github.com/yahuiliu99/PointConT.

Pearl shape classification using deep convolutional neural networks from Tahitian pearl rotation in Pinctada Margaritifera

Pearl shape classification using deep convolutional neural networks from Tahitian pearl rotation in Pinctada Margaritifera

Shape classification using a new shape descriptor and multi-view learning

Shape classification is considered as a vital task in solving many computer vision problems. Different factors such as affine transformations, scaling, rotations, variation in perspective, noise and occlusion have made the shape classification problem to be a hard problem. This work investigates a new shape descriptor that extracts different features from each boundary pixel. This makes it to be more informative and discriminant in comparison with other descriptors. After feature extraction, the “bag-of-features (BoF)” model is employed to construct the final representation for each image. To enhance the functionality of the BoF model, a novel codebook generation approach is presented. The proposed approach tends to derive a more meaningful visual codebook. Consequently, the produced feature vectors can handle inter- and intra-class variations more effectively. Comprehensive experiments conducted on the various complicated shape datasets show the supremacy of our approach compared to other methods.

14th–15th century potter’s workshop in Vilnius, Subačiaus street 11: statistical analysis of the shapes of household pottery

In this article, the shape of 14th–15th century pottery from Subačiaus Street 11, Vilnius, is analyzed through the application of big data analysis techniques (relational data model, Structured Query Language (SQL), programming language Python, and statistical methods). Two main tasks were carried out as part of this study, namely, the classification of the earliest shapes of pottery found in the potter’s workshop in Vilnius and the implementation of IT methods in the research of archaeological household pottery of Vilnius. A total of 1628 fragments of the upper part of vessels and complete vessels from two middens were analyzed to discern a morphological relationship of assemblages, to assess the diversity of shapes, as well as similarities and differences in techniques used during production. Further tasks include the classification of the shape of rims, profiles, and decorations. Both middens were found to contain pairs of similarly shaped sherds, leading us to conclude that although production was characterized by similar features, they were not identical—possibly due to minor differences in the chronology of the assemblages, different circumstances of artifacts’ deposition in archaeological layers, or because the two middens were used by several craftsmen with a common work experience.

Breast morphology classification method based on generalized neural network

In order to establish a breast shape classification method with general indicators, this paper selects six measurement indicators that can characterize the three-dimensional shape of the breast, such as circumference, length, width, and height. The relevant data of the measurement indicators are used to establish a generalized neural network model, and the generalized neural network is applied to the evaluation of breast shape classification recognition. The MATLAB software toolbox is used, based on training sample data, We have established a generalized neural network (GRNN) model, which involves establishing corresponding functional relationships through training to approximate the training data. After obtaining the function relationship through training, the predicted values were obtained, and the prediction results reflected that the established generalized neural network GRNN model is quite reliable for predicting the category of breast shape. By establishing a mapping relationship between the input quantity of breast feature parameters and the output quantity of categories, training samples can be established. When encountering new test data, these data can also be used as test samples. Based on the established model, appropriate spreads can be adjusted to predict the output quantity, providing relevant personnel with basic and simple basis for judgment and evaluation. It can also provide convenience for customers in various consumption channels Accurate bra size recommendations can help consumers purchase bras that fit well and are comfortable.

Pebbles, graphs and equilibria: Higher order shape descriptors for sedimentary particles

AbstractWhile three-dimensional measurement technology is spreading fast, its meaningful application to sedimentary geology still lacks content. Classical shape descriptors (such as axis ratios, circularity of projection) were not inherently three-dimensional, because no such technology existed. Recently a new class of three-dimensional descriptors, collectively referred to as mechanical descriptors, has been introduced and applied for a broad range of sedimentary particles. First-order mechanical descriptors (registered for each pebble as a pair {S, U} of integers), refer to the respective numbers of stable and unstable static equilibria and can be reliably detected by hand experiments. However, they have limited ability of distinction, as the majority of coastal pebbles fall into primary class . Higher-order mechanical descriptors offer a more refined distinction. However, for the extraction of these descriptors (registered as graphs for each pebble), hand measurements are not an option and even computer-based extraction from 3D scans offers a formidable challenge. Here we not only describe and implement an algorithm to perform this task, but also apply it to a collection of 271 pebbles with various lithologies, illustrating that the application of higher-order descriptors is a viable option for geologists. We also show that the so-far uncharted connection between the two known secondary descriptors, the so-called Morse–Smale graph and the Reeb-graph, can be established via a third order descriptor which we call the master graph.

MLGPnet: Multi-granularity neural network for 3D shape recognition using pyramid data

This paper presents a Multi-Granularity 3D shape recognition network comprising point-granularity, line-granularity, and Pyramid-granularity networks, as well as multi-granularity convolutional layers (MLGPnet). The network takes pyramid data with high-level features generated from mesh data as input. The point-granularity, line-granularity, and pyramid-granularity networks respectively generate features at the point, line, and pyramid levels. Finally, two multi-granularity convolutional layers merge the features from these different levels to generate more efficient 3D shape global features. Compared to some classical 3D shape recognition network models, the proposed network achieves superior results on three publicly general-purpose datasets. Notably, among all mesh-based recognition networks, the proposed network demonstrates the best recognition accuracy and retrieval rate. Furthermore, the proposed network model performs better in terms of training time and model complexity, with faster training time and fewer model parameters, resulting in faster recognition speed.

Numerical investigation of the thermo-hydraulic performance of a shark denticle-inspired plate fin heat exchanger

Growing demand for increased power dissipation is fuelling the need for the design of more efficient heat exchangers. Modifying fin geometry is an effective way of improving heat transfer efficiency and of reducing flow resistance for plate fin heat exchangers. To date, fin designs have mostly revolved around classical shapes such as pins, ellipses, and rectangles, due to limitations of conventional manufacturing technologies. However, with recent advancements in metal additive manufacturing, there is an opportunity of redesigning approaches and ways of thinking to create novel fin geometries. In this study, biomimicry was used as a tool to reverse engineer a novel geometry for a plate fin heat exchanger based on a denticle, which is a mesoscopic structure on the skin of sharks. The shape of the denticle is streamlined; therefore, showing the possibility of enhancing fluid-to-solid contact if used as geometry for a plate fin heat exchanger. The thermo-hydraulic performance of the denticle was evaluated with respect to a rectangular, cylindrical and an elliptical fin (NACA 0030), using conjugate-heat-transfer simulations on ANSYS-Fluent. The numerical model was calibrated and validated based on experimental results from an additively manufactured rectangular plate fin heat exchanger. Results demonstrated that over the range of tested Reynolds Numbers, 9.8x104 < Re ≤ 22.9x104, the denticle fin had average Nusselt number improvements of 13.1 %, 5.4 % and 75.7 % with respect to the rectangular fin, the NACA 0030 and cylindrical fin, respectively. At Re = 22.9x104, a 26 % decrease in pressure drop was noted for the denticle with respect to the rectangular fin, with maximum thermal performance factors (ƞ) of 1.29, 0.81 and 1.53 as noted for the denticle fin with respect to the rectangular fin, NACA 0030 fin and cylindrical fin, respectively.

Automated and reliable detection of multi-diseases on chest X-ray images using optimized ensemble transfer learning

Chest radiography is a comparatively low-cost, most probably applied in a medical process that brings crucial information to discover diagnostic conclusions. Chest X-rays (CXR) are used in the prognosis of chest diseases that includes asthma, lung cancer, pneumonia, and COVID-19. Artificial intelligence detects the multi-disease automatically to enhance its proficiency and performance by solving image detection complications with various machine learning and deep learning approaches. Convolutional Neural Network (CNN) has been designed for the advancement of computerized recognition systems. For the classification of medical images, texture, shape, size, and tissue constitution are essential features for detecting diseases. Hence, huge input features are merged with deep CNN validated to increase the effectiveness of CXR analysis. In addition, multi-scale features are visualized in CNN for the detection of variable sizes of thoracic diseases. This paper has implemented a novel multi-disease diagnosis model using chest X-ray images through deep learning approaches, and thus, it helps to minimize the computational cost. The standard CXR images are gathered from the standard datasets. After, the pre-processing is performed for cleaning and contrast the images. Further, the image segmentation is carried out using Optimized DeepLabv3 (ODeepLabv3), where the optimization of parameters in DeepLabv3 is done by a new Mutation Rate-based Lion Algorithm (MR-LA). Then, multi-disease classification is carried out through Optimized Ensemble Transfer Learning (OETL), where the OETL is carried out through VGG16, ResNet, ImageNet, MobileNet, GoogleNet, Inception, and Xception. Here, the parameter optimization of all models is done by the same MR-LA. The proposed model improved the effectiveness and performance in expressions of sensitivity, precision, and specificity factors and get higher classification and detection accuracy. The outcome of the recommended method over other models is identified by the relative analysis conducted.

Classification of lower limb body shapes of paralysed female wheelchair users in Hebei, China: analysis of anthropometric data using clustering approach

Lower limb body shape is important in the design of functional pants. The skin, muscles, and body shapes of the lower limbs of wheelchair users may differ from healthy people because of the different shapes of their legs and the prolonged seating position. This study aimed to classify the shapes of the lower limbs of adult female wheelchair users. The lower body measurement of 384 female wheelchair users was obtained. The principal component analysis and two-step cluster analysis were used to categorise the body shapes into three different types and five different size standards. Based on the study findings, female wheelchairs have larger waist, belly, and hip circumferences than healthy individuals, with 89.3% of them having prominent hips. Therefore, the design and production of trousers for wheelchair users should take into consideration the classification of lower limb shapes and sizes reported in this study. Practitioner summary: This work initiated the investigation of human body size assessment of clothes for handicapped persons in China, allowing paraplegic female wheelchair users to wear adapted trousers.

On geometric shape construction via growth operations

We study algorithmic growth processes under a geometric setting. Each process begins with an initial shape of nodes SI=S0 and, in every time step t≥1, by applying (in parallel) one or more growth operations of a specific type to the current shape, St−1, generates the next, St, always satisfying |St|>|St−1|. We define three types of growth operations and explore the algorithmic and structural properties of their resulting processes. Our goal is to characterize the classes of shapes that can be constructed in O(log⁡n) or polylog n time steps, n being the size of the final shape SF. Moreover, we want to determine whether a given shape SF can be constructed from a given initial shape SI using a finite sequence of growth operations of a given type, called a constructor ofSF. We give exact and partial characterizations of classes of shapes that can be constructed in polylog n time steps, polynomial-time centralized algorithms for deciding reachability between pairs of input shapes (SI,SF) and for generating constructors when SF can be constructed from SI, as well as some negative results.

Intelligent image-based deformability assessment of red blood cells via dynamic shape classification

Red blood cells (RBC), the primary component of whole blood, play an essential role in maintaining blood fluidity and mass transport throughout the human body. RBC deformability is an important property that indicates RBC quality and diseases. Among the techniques for RBC deformability measurements, the hydrodynamic image-based approach represents the most advanced method. However, RBC have a complex shape and high deformability, they display a wide range of shapes during the transition in microvessels, making it a challenge for deformation measurement. Therefore, we propose an image-based RBC deformability assessment via a shape-classification approach (IRIS) to solve this issue, where six shapes are identified by deep learning technique. Then, the deformations of RBC stiffened by different glutaraldehyde (GA) concentrations are measured and compared. The results elucidate that IRIS is able to sensitively distinguish 0.001% GA treated RBC from the control. Moreover, it is superior to two-shape classification or zero-shape classification in terms of sensitivity for deformability quantification. IRIS demonstrates an improved sensitivity for deformability analysis, high automaticity and unprecedented large data volume by applying deep learning method, it has a vast potential to benefit the hemorheology field as a powerful tool for early detection and diagnosis of RBC related diseases.

Automated rapid artefact surface area measurement from imagery with computer vision

Automated surface area measurements have been of interest to archaeologists since digital imagery began allowing researchers to remotely collect artefact metrics. We present a method of automatically measuring 2D surface area from artefact planform images employing computer vision in Python. The Python script, provided as a .py file in supplementary data, creates boundaries around regions of relatively homogeneous pixels (artefacts) in the image. These bounded regions are called contours. A count of the number of pixels within each contour provides a surface area in pixels. A circular reference object provides a conversion factor for the contours, as well as a point of reference for geometric accuracy of outputs.&#x0D; Measurements of 2D artefact surface area can be used in combination with measurements of length, width, thickness, and mass, or in some cases, replace such measurements. As presented, this technique provides utility to archaeology with applications to new documentation of artefacts, archived artefact images containing a scale, as well as landscape geoarchaeology and sedimentary contexts. Limitations of this type of surface area measurement include the requirement of the image background being of a solid colour heavily contrasting the artefacts being measured. Effectively, the background requirement limits deployment supporting collection of rapid field measurements from in-situ surface scatters without modification to the script or manipulation of the artefacts. Analytical applications utilizing this technique include studies of relative artefact abundance, shape and size class characterizations in artefact scatters, and redistribution of artefacts by geomorphological processes.

Learning from vector data: enhancing vector-based shape encoding and shape classification for map generalization purposes

ABSTRACT Map generalization is a complex task that requires a high level of spatial cognition, and deep learning techniques have shown in numerous research fields that they could match or even outplay human cognition when knowledge is implicitly in the data. First experiments that apply deep learning techniques to map generalization tasks thereby adapt models from image processing, creating input data by rasterizing spatial vector data. Because image-based learning has major shortcomings for map generalization, this article investigates possibilities to learn directly from vector data, utilizing vector-based encoding schemes. First, we enhance preprocessing methods to match essential properties of deep learning models – namely regularity and feature description – and evaluate the performance of Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Graph Convolutional Neural Networks (GCNN) in combination with a feature-based encoding scheme. The results show that feature descriptors improve the accuracy of all three neural networks, and that the overall performances of the models are quite similar for both polygon and polyline shape classification tasks. In a second step, we implement an exemplary building generalization workflow based on shape classification and template matching, and discuss the generalization results based on a case study.