Global Shape Representation Research Articles

Isolating Global Form Processing Using Shape Metamers

A critical disconnect between human and machine vision is that people have a remarkable ability to represent and understand the shapes of objects, whereas deep convolutional neural networks (CNNs) operate more as local texture analyzers with no propensity to perceive global form. To isolate and study the representation of global form, we developed a novel stimulus manipulation which enables us to probe global form similarity in a way that is completely isolated from local similarity. Specifically, our method renders natural images as a composition of Gabor elements of different orientations and scales. From this basis, we generate shape metamers--image pairs that differ at each local gabor, but share the same global form information and are nearly perceptually indistinguishable by humans. And, we generate anti-metamers — image pairs that differ by the same amount locally, but disrupt global form information, and look nothing alike to humans. Leveraging these stimuli as a litmus test for global form perception, we find that CNNs trained to do object categorization show little sensitivity to global form information in their feature spaces, which encode both the shape metamers and anti-metamers as equally similar. However, we find that both vision transformer models with self-attention layers, as well as our simpler custom models with hand-designed “association fields,” can learn longer-range relationships between local features, and have increased sensitivity to global form. These findings highlight that the CNN model lacks sufficient inductive biases to learn global form information, and that self-attention and association field mechanisms may serve as key precursor operations which amplify relevant local features. We propose this multiplicative operation is critical to enable downstream mechanisms to encode relationships between primarily shape-defining local features, en route to a more explicit global shape representation.

DcTr: Noise-robust point cloud completion by dual-channel transformer with cross-attention

Current point cloud completion research mainly utilizes the global shape representation and local features to recover the missing regions of 3D shape for the partial point cloud. However, these methods suffer from inefficient utilization of local features and unstructured points prediction in local patches, hardly resulting in a well-arranged structure for points. To tackle these problems, we propose to employ Dual-channel Transformer and Cross-attention (CA) for point cloud completion (DcTr). The DcTr is apt at using local features and preserving a well-structured generation process. Specifically, the dual-channel transformer leverages point-wise attention and channel-wise attention to summarize the deconvolution patterns used in the previous Dual-channel Transformer Point Deconvolution (DCTPD) stage to produce the deconvolution in the current DCTPD stage. Meanwhile, we employ cross-attention to convey the geometric information from the local regions of incomplete point clouds for the generation of complete ones at different resolutions. In this way, we can generate the locally compact and structured point cloud by capturing the structure characteristic of 3D shape in local patches. Our experimental results indicate that DcTr outperforms the state-of-the-art point cloud completion methods under several benchmarks and is robust to various kinds of noise.

The developmental trajectories of children's reorientation to global and local properties of environmental geometry.

The way in which organisms represent the shape of their environments during navigation has been debated in cognitive, comparative, and developmental psychology. While there is evidence that adult humans encode the entire boundary shape of an environment (a global-shape representation), there are also data demonstrating that organisms reorient using only segments of the boundary that signal a goal location (a local-shape representation). Developmental studies offer unique insights into this debate; however, most studies have used designs that cannot dissociate the type of boundary-shape representation that children use to guide reorientation. Thus, we examined the developmental trajectories of children's reorientation according to local and global boundary shape. Participants aged 6-12 years were trained to find a goal hidden in one corner of a virtual arena, after which they were required to reorient in a novel test arena. From 10.5 years, children performed above chance when the test arena permitted reorientation based only on local-shape (Experiment 2), or only global-shape (Experiment 3) information. Moreover, when these responses were placed into conflict, older children reoriented with respect to global-shape information (Experiment 4). These age-related findings were not due to older children being better able to reorient in virtual environments per se: when trained and tested within the same environment (Experiment 1), children performed above chance from 6 years. Together, our results suggest (a) the ability to reorient on the basis of global- and local-shape representations develops in parallel, and (b) shape-based information is weighted to determine which representation informs reorientation. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Feedback from lateral occipital cortex to V1/V2 triggers object completion: Evidence from functional magnetic resonance imaging and dynamic causal modeling.

Illusory figures demonstrate the visual system's ability to integrate disparate parts into coherent wholes. We probed this object integration process by either presenting an integrated diamond shape or a comparable ungrouped configuration that did not render a complete object. Two tasks were used that either required localization of a target dot (relative to the presented configuration) or discrimination of the dot's luminance. The results showed that only when the configuration was task relevant (in the localization task), performance benefited from the presentation of an integrated object. Concurrent functional magnetic resonance imaging was performed and analyzed using dynamic causal modeling to investigate the (causal) relationship between regions that are associated with illusory figure completion. We found object‐specific feedback connections between the lateral occipital cortex (LOC) and early visual cortex (V1/V2). These modulatory connections persisted across task demands and hemispheres. Our results thus provide direct evidence that interactions between mid‐level and early visual processing regions engage in illusory figure perception. These data suggest that LOC first integrates inputs from multiple neurons in lower‐level cortices, generating a global shape representation while more fine‐graded object details are then determined via feedback to early visual areas, independently of the current task demands.

Predict Alzheimer\u2019s disease using hippocampus MRI data: a lightweight 3D deep convolutional network model with visual and global shape representations

BackgroundAlzheimer’s disease (AD) is a progressive and irreversible brain disorder. Hippocampus is one of the involved regions and its atrophy is a widely used biomarker for AD diagnosis. We have recently developed DenseCNN, a lightweight 3D deep convolutional network model, for AD classification based on hippocampus magnetic resonance imaging (MRI) segments. In addition to the visual features of the hippocampus segments, the global shape representations of the hippocampus are also important for AD diagnosis. In this study, we propose DenseCNN2, a deep convolutional network model for AD classification by incorporating global shape representations along with hippocampus segmentations.MethodsThe data was obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and was T1-weighted structural MRI from initial screening or baseline, including ADNI 1,2/GO and 3. DenseCNN2 was trained and evaluated with 326 AD subjects and 607 CN hippocampus MRI using 5-fold cross-validation strategy. DenseCNN2 was compared with other state-of-the-art machine learning approaches for the task of AD classification.ResultsWe showed that DenseCNN2 with combined visual and global shape features performed better than deep learning models with visual or global shape features alone. DenseCNN2 achieved an average accuracy of 0.925, sensitivity of 0.882, specificity of 0.949, and area under curve (AUC) of 0.978, which are better than or comparable to the state-of-the-art methods in AD classification. Data visualization analysis through 2D embedding of UMAP confirmed that global shape features improved class discrimination between AD and normal.ConclusionDenseCNN2, a lightweight 3D deep convolutional network model based on combined hippocampus segmentations and global shape features, achieved high performance and has potential as an efficient diagnostic tool for AD classification.

ASHF-Net: Adaptive Sampling and Hierarchical Folding Network for Robust Point Cloud Completion

Estimating the complete 3D point cloud from an incomplete one lies at the core of many vision and robotics applications. Existing methods typically predict the complete point cloud based on the global shape representation extracted from the incomplete input. Although they could predict the overall shape of 3D objects, they are incapable of generating structure details of objects. Moreover, the partial input point sets obtained from range scans are often sparse, noisy and non-uniform, which largely hinder shape completion. In this paper, we propose an adaptive sampling and hierarchical folding network (ASHF-Net) for robust 3D point cloud completion. Our main contributions are two-fold. First, we propose a denoising auto-encoder with an adaptive sampling module, aiming at learning robust local region features that are insensitive to noise. Second, we propose a hierarchical folding decoder with the gated skip-attention and multi-resolution completion goal to effectively exploit the local structure details of partial inputs. We also design a KL regularization term to evenly distribute the generated points. Extensive experiments demonstrate that our method outperforms existing state-of-the-art methods on multiple 3D point cloud completion benchmarks.

Fine-Grained 3D Shape Classification With Hierarchical Part-View Attention.

Fine-grained 3D shape classification is important for shape understanding and analysis, which poses a challenging research problem. However, the studies on the fine-grained 3D shape classification have rarely been explored, due to the lack of fine-grained 3D shape benchmarks. To address this issue, we first introduce a new 3D shape dataset (named FG3D dataset) with fine-grained class labels, which consists of three categories including airplane, car and chair. Each category consists of several subcategories at a fine-grained level. According to our experiments under this fine-grained dataset, we find that state-of-the-art methods are significantly limited by the small variance among subcategories in the same category. To resolve this problem, we further propose a novel fine-grained 3D shape classification method named FG3D-Net to capture the fine-grained local details of 3D shapes from multiple rendered views. Specifically, we first train a Region Proposal Network (RPN) to detect the generally semantic parts inside multiple views under the benchmark of generally semantic part detection. Then, we design a hierarchical part-view attention aggregation module to learn a global shape representation by aggregating generally semantic part features, which preserves the local details of 3D shapes. The part-view attention module hierarchically leverages part-level and view-level attention to increase the discriminability of our features. The part-level attention highlights the important parts in each view while the view-level attention highlights the discriminative views among all the views of the same object. In addition, we integrate a Recurrent Neural Network (RNN) to capture the spatial relationships among sequential views from different viewpoints. Our results under the fine-grained 3D shape dataset show that our method outperforms other state-of-the-art methods. The FG3D dataset is available at https://github.com/liuxinhai/FG3D-Net.

Content-Based Image Retrieval Features: A Survey

Content-Based Image Retrieval (CBIR) systems have been used for the searching of relevant images in various research areas. In CBIR systems features such as shape, texture and color are used. The extraction of features is the main step on which the retrieval results depend. Color features in CBIR are used as in the color histogram, color moments, conventional color correlogram and color histogram. Color space selection is used to represent the information of color of the pixels of the query image. The shape is the basic characteristic of segmented regions of an image. Different methods are introduced for better retrieval using different shape representation techniques; earlier the global shape representations were used but with time moved towards local shape representations. The local shape is more related to the expressing of result instead of the method. Local shape features may be derived from the texture properties and the color derivatives. Texture features have been used for images of documents, segmentation-based recognition,and satellite images. Texture features are used in different CBIR systems along with color, shape, geometrical structure and sift features.

The maintenance and updating of representations of no longer visible objects and their parts.

When an object partially or completely disappears behind an occluding surface, a representation of that object persists. For example, fragments of no longer visible objects can serve as an input into mid-level constructive visual processes, interacting and integrating with currently visible portions to form perceptual units and global motion signals. Remarkably, these persistent representations need not be static and can have their positions and orientations updated postdictively as new information becomes visible. In this chapter, we highlight historical considerations, behavioral evidence, and neural correlates of this type of representational updating of no longer visible information at three distinct levels of visual processing. At the lowest level, we discuss spatiotemporal boundary formation in which visual transients can be integrated over space and time to construct local illusory edges, global form, and global motion percepts. At an intermediate level, we review how the visual system updates form information seen at one moment in time and integrates it with subsequently available information to generate global shape and motion representations (e.g., spatiotemporal form integration and anorthoscopic perception). At a higher level, when an entire object completely disappears behind an occluder, the object's identity and predicted position can be maintained in the absence of visual information.

Learning a discriminative deformation-invariant 3D shape descriptor via many-to-one encoder

Recent advances in 3D acquisition techniques have led to a rapid increase in the size of database of three dimensional (3D) models across areas as diverse as engineering, medicine and biology, etc. Therefore, developing an efficient shape retrieval method has been attracting more and more attention in recent years. In this paper, we have developed a novel learning paradigm for extracting a concise data-driven shape descriptor to address challenging issues posed by structural deformation variations and noise present in 3D models. First, we use the scale invariant heat kernel signature (SIHKS) to describe the vertex of the shape. The locality-constrained linear coding (LLC) is employed to encode each vertex of the shape to form the global shape representation. Then we develop a discriminative shape descriptor for retrieval using many-to-one encoder. Our proposed shape descriptor is extensively evaluated on three well-known benchmark datasets including McGill, SHREC’10 ShapeGoogle and SHREC’14 human. Experimental results on 3D shape retrieval demonstrate the superior performance of our proposed method over the state-of-the-art methods and is robust to large deformations.

Dynamic systems in human face recognition: A novel face processing model

Event Abstract Back to Event Dynamic systems in human face recognition: A novel face processing model William Comfort1* and Yossi Zana1 1 Universidade Federal do ABC, Centre for Mathematics, Computation and Cognition, Brazil Face recognition is characterised by the interaction of a distributed network of discrete brain regions and by consecutive stages of rapid and efficient detection and individuation. We examine the evidence for these stages and synthesise current findings in a novel model which accounts for the differential task demands of categorising an object as a face out of a series of object categories, and for categorising a face at the individual level. The model functionally segregates face processing according to three main aspects: (a) the time course of face processing; (b) the view dependence of face-selective processing; (c) the differential response to familiar or unfamiliar faces across face-selective regions. Early visual information from V1 is conveyed to the occipital face area (OFA), where it is coded and identified in terms of facial features. A parallel stream of information is directed to the lateral occipital complex (LOC), where a global shape representation of the face is computed via the integration of multiple visual cues. A distinct face-sensitive region in the fusiform gyrus (FG) responds to both face parts and face configuration, indicating a unified representation of the structure of a face. The identification of an object as a face appears to occur via interaction between the FG, OFA and LOC. Recognition of facial identity relies on the integration of a view-dependent representation of facial identity based on the current perception, with a view-invariant model derived from semantic knowledge. This integration appears to occur via bidirectional interaction between the anterior temporal cortex and medial to anterior FG. Additional connections to frontal regions facilitate the retrieval of person-related semantic knowledge as well as the modulation of visual attention to familiar faces. We discuss recent TMS and fMRI pattern mapping results in the context of our proposed model focusing on separate, autonomous stages of face detection and individuation. Keywords: viewpoint invariance, face recognition, Cognitive Modelling, Face detection, familiar face processing Conference: XII International Conference on Cognitive Neuroscience (ICON-XII), Brisbane, Queensland, Australia, 27 Jul - 31 Jul, 2014. Presentation Type: Poster Topic: Sensation and Perception Citation: Comfort W and Zana Y (2015). Dynamic systems in human face recognition: A novel face processing model. Conference Abstract: XII International Conference on Cognitive Neuroscience (ICON-XII). doi: 10.3389/conf.fnhum.2015.217.00179 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 19 Feb 2015; Published Online: 24 Apr 2015. * Correspondence: Mr. William Comfort, Universidade Federal do ABC, Centre for Mathematics, Computation and Cognition, São Bernardo, Brazil, wecomfort@gmail.com Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers William Comfort Yossi Zana Google William Comfort Yossi Zana Google Scholar William Comfort Yossi Zana PubMed William Comfort Yossi Zana Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.

Stable geodesic surface signatures

Shape analysis plays a fundamental role in computer graphics. We present a novel global and intrinsic shape representation for shape analysis, called stable geodesic signature. It is based on the theory of stable closed geodesics and surface Ricci flow. We examine the surface by dynamically deforming it by metric design through Ricci flow, then we observe the behavior of the stable closed geodesics under the evolving Riemannian metrics. When a metric is deforming, some stable geodesic loops will become unstable and shrink to points, or some geodesic loops may merge. The number of stable geodesics forms the signature, which is general for arbitrary surfaces. Experiments on a large amount of surfaces demonstrate the efficiency and efficacy of the stable geodesic signature for shape analysis.

Shape-Based Object Detection via Boundary Structure Segmentation

We address the problem of object detection and segmentation using global holistic properties of object shape. Global shape representations are highly susceptible to clutter inevitably present in realistic images, and thus can be applied robustly only using a precise segmentation of the object. To this end, we propose a figure/ground segmentation method for extraction of image regions that resemble the global properties of a model boundary structure and are perceptually salient. Our shape representation, called the chordiogram, is based on geometric relationships of object boundary edges, while the perceptual saliency cues we use favor coherent regions distinct from the background. We formulate the segmentation problem as an integer quadratic program and use a semidefinite programming relaxation to solve it. The obtained solutions provide a segmentation of the object as well as a detection score used for object recognition. Our single-step approach achieves state-of-the-art performance on several object detection and segmentation benchmarks.

Scale invariant and deformation tolerant partial shape matching

We present a novel approach to the problem of establishing the best match between an open contour and a part of a closed contour. At the heart of the proposed scheme lies a novel shape descriptor that also permits the quantification of local scale. Shape descriptors are computed along open or closed contours in a spatially non-uniform manner. The resulting ordered collections of shape descriptors constitute the global shape representation. A variant of an existing Dynamic Time Warping (DTW) matching technique is proposed to handle the matching of shape representations. Due to the properties of the employed shape descriptor, sampling scheme and matching procedure, the proposed approach performs partial shape matching that is invariant to Euclidean transformations, starting point as well as to considerable shape deformations. Additionally, the problem of matching closed-to-closed contours is naturally treated as a special case. Extensive experiments on benchmark datasets but also in the context of specific applications, demonstrate that the proposed scheme outperforms existing methods for the problem of partial shape matching and performs comparably to methods for full shape matching.

Intrinsic Geometric Scale Space by Shape Diffusion

This paper formalizes a novel, intrinsic geometric scale space (IGSS) of 3D surface shapes. The intrinsic geometry of a surface is diffused by means of the Ricci flow for the generation of a geometric scale space. We rigorously prove that this multiscale shape representation satisfies the axiomatic causality property. Within the theoretical framework, we further present a feature-based shape representation derived from IGSS processing, which is shown to be theoretically plausible and practically effective. By integrating the concept of scale-dependent saliency into the shape description, this representation is not only highly descriptive of the local structures, but also exhibits several desired characteristics of global shape representations, such as being compact, robust to noise and computationally efficient. We demonstrate the capabilities of our approach through salient geometric feature detection and highly discriminative matching of 3D scans.

Fast protein tertiary structure retrieval based on global surface shape similarity

Characterization and identification of similar tertiary structure of proteins provides rich information for investigating function and evolution. The importance of structure similarity searches is increasing as structure databases continue to expand, partly due to the structural genomics projects. A crucial drawback of conventional protein structure comparison methods, which compare structures by their main-chain orientation or the spatial arrangement of secondary structure, is that a database search is too slow to be done in real-time. Here we introduce a global surface shape representation by three-dimensional (3D) Zernike descriptors, which represent a protein structure compactly as a series expansion of 3D functions. With this simplified representation, the search speed against a few thousand structures takes less than a minute. To investigate the agreement between surface representation defined by 3D Zernike descriptor and conventional main-chain based representation, a benchmark was performed against a protein classification generated by the combinatorial extension algorithm. Despite the different representation, 3D Zernike descriptor retrieved proteins of the same conformation defined by combinatorial extension in 89.6% of the cases within the top five closest structures. The real-time protein structure search by 3D Zernike descriptor will open up new possibility of large-scale global and local protein surface shape comparison.

Neurophysiological Evidence for the Time Course of Activation of Global Shape, Part, and Local Contour Representations during Visual Object Categorization and Memory

Categorization of visual objects entails matching a percept to long-term representations of structural knowledge. This object model selection is central to theories of human visual cognition, but the representational format(s) is largely unknown. To characterize these neural representations, event-related brain potentials (ERPs) to fragmented objects during an indirect memory test were compared when only local contour features, but not global shapes of the object and its parts, differed between encoding and retrieval experiences. The ERP effects revealed that the format of object representations varies across time according to the particular neural processing and memory system currently engaged. An occipito-temporal P2(00) showed implicit memory modulation to items that repeatedly engaged similar perceptual grouping processes but not items that merely reinstantiated visual features. After 500 msec, memory modulation of a late positive complex, indexing secondary categorization and/or explicit recollection processes, was sensitive to local contour changes. In between, a frontocentral N350, indexing the model selection and an implicit perceptual memory system, showed reactivation of object representations whenever the same global shapes were reactivated, despite local feature differences. These and prior N350 findings provide direct neurophysiological evidence that the neural representations supporting object categorization include knowledge beyond local contours and about higher-order perceptual structures, such as the global shapes of the object and its parts, that can differ between object views. The N350 is proposed to index a second state of interactive, recurrent, and feedback processing in occipital and ventral temporal neocortex supporting higher-order cognitive abilities and phenomenological awareness with objects.

Perceived orientation of complex shape reflects graded part decomposition

Although the orientation of line segments and simple shapes is a well-studied area of vision, little is known about geometric factors that influence perceived orientation of complex multipart shapes. The study of these factors is of interest because it allows for an insight into the basic problem of how local geometric attributes are integrated perceptually into a global shape representation. We examined the perceived orientation of two-part shapes using an adjustment method and a 2AFC task. In particular, we investigated the influence of the perceptual salience, or distinctiveness, of a part--as defined by the turning angles at its boundaries--and its area relative to the main "base" part. In contrast to previous results on simple shapes, our results exhibited large and systematic deviations of perceived orientation from the principal axis of the shape. For shapes with sharp part boundaries, perceived global orientation deviated maximally from the principal axis and was approximated by the axis of the main base part of the shape. With weakening part boundaries, the perceived orientation gradually approached the principal axis of the entire shape, reflecting that both parts were taken into account in estimating orientation. The results are consistent with a differentially weighted principal-axis computation in which the attached part is given systematically lower weighting with increasing turning angles at the part boundaries. They thus allow a quantitative characterization of part salience in terms of part independence: Turning angles at a part's boundaries determine the extent to which its influence is perceptually separable from the rest of the shape. We suggest that Robust Statistics may provide a useful framework for quantifying the influence of part segmentation on visual estimation.