- 3D Shape Modeling and Analysis
- Image Processing and 3D Reconstruction
- Computer Graphics and Visualization Techniques
- 3D Surveying and Cultural Heritage
- Image Retrieval and Classification Techniques
- Medical Image Segmentation Techniques
- Advanced Image and Video Retrieval Techniques
- Computational Geometry and Mesh Generation
- Color perception and design
- Advanced Vision and Imaging
- Human Pose and Action Recognition
- Color Science and Applications
- Morphological variations and asymmetry
- Video Analysis and Summarization
- Advanced Numerical Analysis Techniques
- Aesthetic Perception and Analysis
- Multimodal Machine Learning Applications
- Human Motion and Animation
- Graph Theory and Algorithms
- Emotion and Mood Recognition
- Artistic and Creative Research
- Digital Media and Visual Art
- Medical Imaging Techniques and Applications
- Reinforcement Learning in Robotics
- Additive Manufacturing and 3D Printing Technologies
Carleton University
2016-2025
Shenzhen University
2021
Software (Spain)
2021
University of Konstanz
2021
Tel Aviv University
2014-2015
Simon Fraser University
2006-2013
Abstract We review methods designed to compute correspondences between geometric shapes represented by triangle meshes, contours or point sets. This survey is motivated in part recent developments space–time registration, where one seeks a correspondence non‐rigid and time‐varying surfaces, semantic shape analysis, which underlines trend incorporate understanding into the analysis pipeline. Establishing meaningful often difficult because it generally requires an of structure at both local...
Unsupervised co-analysis of a set shapes is difficult problem since the geometry alone cannot always fully describe semantics shape parts. In this paper, we propose semi-supervised learning method where user actively assists in by iteratively providing inputs that progressively constrain system. We introduce novel constrained clustering based on spring system which embeds elements to better respect their inter-distances feature space together with user-given constraints. also present an...
We introduce a learning framework for automated floorplan generation which combines generative modeling using deep neural networks and user-in-the-loop designs to enable human users provide sparse design constraints. Such constraints are represented by layout graph. The core component of our is network, Graph2Plan, converts graph, along with building boundary, into that fulfills both the boundary Given an input we allow user specify room counts other constraints, used retrieve set...
We introduce an algorithm for unsupervised co-segmentation of a set shapes so as to reveal the semantic shape parts and establish their correspondence across set. The input may exhibit significant variability where do not admit proper spatial alignment corresponding in any pair be geometrically dissimilar. Our can handle such challenging sets since, first, we perform co-analysis descriptor space, combination descriptors relates independently pose, location, cardinality. Secondly, exploit key...
Abstract Spectral methods for mesh processing and analysis rely on the eigenvalues, eigenvectors, or eigenspace projections derived from appropriately defined operators to carry out desired tasks. Early work in this area can be traced back seminal paper by Taubin 1995, where spectral of geometry based a combinatorial Laplacian aids our understanding low‐pass filtering approach smoothing. Over past 15 years, list applications which utilize eigenstructures variety different manners have been...
Abstract Non‐rigid 3D shape correspondence is a fundamental and difficult problem. Most applications which require rely on manually selected markers. Without user assistance, the performances of existing automatic methods depend strongly good initial alignment or prior, they generally do not tolerate large variations. We present an feature algorithm capable handling large, non‐rigid variations, as well partial matching. This made possible by leveraging power state‐of‐the‐art mesh deformation...
We introduce an algorithm for unsupervised co-segmentation of a set shapes so as to reveal the semantic shape parts and establish their correspondence across set. The input may exhibit significant variability where do not admit proper spatial alignment corresponding in any pair be geometrically dissimilar. Our can handle such challenging sets since, first, we perform co-analysis descriptor space , combination descriptors relates independently pose, location, cardinality. Secondly, exploit...
We present a shape segmentation method for complete and incomplete shapes. The key idea is to directly optimize the decomposition based on characterization of expected geometry part in shape. Rather than setting number parts advance, we search smallest that admit geometric parts. an intermediate-level analysis, where first decomposed into approximate convex components, which are then merged consistent nonlocal signature. Our designed handle shapes, represented by point clouds. show results...
We present an algorithm for finding a meaningful vertex-to-vertex correspondence between two triangle meshes, which is designed to handle general non-rigid transformations. Our operates on embeddings of the shapes in spectral domain so as normalize them with respect uniform scaling and rigid-body transformation. Invari-ance shape bending achieved by relying approximate geodesic point proximities mesh capture its shape. To deal moderate stretching, we first raise issue “eigenmode switching”...
We introduce a meta-representation that represents the essence of family shapes. The learns configurations shape parts are common across family, and encapsulates this knowledge with system geometric distributions encode relative arrangements parts. Thus, instead predefined priors, what characterizes is directly learned from set input constructed co-segmented shapes known correspondence. It can then be used in several applications where we seek to preserve identity as members family....
Abstract Classical approaches to shape correspondence base their computation purely on the properties, in particular geometric similarity, of shapes question. Their performance still falls far short that humans challenging cases where corresponding parts may differ significantly geometry or even topology. We stipulate these cases, by involves recognition prior knowledge would play a more dominant role than similarity. introduce an approach part which incorporates imparted training set...
We introduce an unsupervised co-hierarchical analysis of a set shapes, aimed at discovering their hierarchical part structures and revealing relations between geometrically dissimilar yet functionally equivalent shape parts across the set. The core problem is that representative co-selection . For each in set, one hierarchy (tree) selected from among many possible interpretations structure shape. Collectively, tree representatives maximize within-cluster structural similarity them. develop...
Spectral methods for mesh processing and analysis rely on the eigenvalues, eigenvectors, or eigenspace projections derived from appropriately defined operators to carry out desired tasks. Early works in this area can be traced back seminal paper by Taubin 1995, where spectral of geometry based a combinatorial Laplacian aids our understanding low-pass filtering approach smoothing. Over past ten years so, list applications which utilize eigenstructures variety different manners have been...
We introduce a method for co-locating style-defining elements over set of 3D shapes. Our goal is to translate high-level style descriptions, such as “Ming” or “European” furniture models, into explicit and localized regions the geometric models that characterize each style. For style, defined union all are able discriminate Another property they frequently occurring, reflecting shape characteristics appear across multiple shapes same Given an input spanning categories styles, where grouped...
We introduce a contextual descriptor which aims to provide geometric description of the functionality 3D object in context given scene. Differently from previous works, we do not regard as an abstract label or represent it implicitly through agent. Our descriptor, called interaction ICON for short, explicitly represents geometry object-to-object interactions. approach analysis is based on key premise that should mainly be derived interactions between objects and isolation. Specifically,...
We introduce a co-analysis method which learns functionality model for an object category, e.g., strollers or backpacks. Like previous works on functionality, we analyze object-to-object interactions and intra-object properties relations. Differently from works, our goes beyond providing functionality-oriented descriptor single object; it prototypes the of category 3D objects by co-analyzing typical involving category. Furthermore, localizes studied to specific locations, surface patches,...
We study a new and elegant instance of geometric dissection 2D shapes: reversible hinged dissection, which corresponds to dual transform between two shapes where one them can be dissected in its interior then inverted inside-out , with hinges on the shape boundary, reproduce other shape, vice versa. call such or RIOT. Since it is rare for possess even rough RIOT, let alone an exact one, we develop both RIOT construction algorithm quick filtering mechanism pick, from collection, potential...
Abstract Feature‐driven analysis forms the basis of many shape processing tasks, where detected feature points are characterized by local descriptors. Such descriptors have so far been defined to capture regions interest centred at individual points. Using such compare can be problematic when performing partial matching, because region is typically as an isotropic neighbourhood around a point, which does not adapt geometry parts. We introduce bilateral map, descriptor whose two Compared...
We introduce a method for learning model the mobility of parts in 3D objects. Our allows not only to understand dynamic functionalities one or more object, but also apply functions static models. Specifically, learned part can predict mobilities object given form single snapshot reflecting spatial configuration space, and transfer from relevant units training data. The data consists set different motion types. Each unit is composed pair (one moving reference part), along with usage examples...
Abstract A central goal of computer graphics is to provide tools for designing and simulating real or imagined artifacts. An understanding functionality important in enabling such modeling tools. Given that the majority man‐made artifacts are designed serve a certain function, objects often reflected by their geometry, way they organized an environment, interaction with other agents. Thus, recent years, variety methods shape analysis have been developed extract functional information about...
We present a method for data sampling in scatterplots by jointly optimizing point selection different views or classes. Our uses space-filling curves (Z-order curves) that partition set into subsets that, when covered each one sample, provide coreset with good approximation guarantees relation to the original set. For scatterplot matrices multiple views, curves, leading partitions of given multi-class scatterplots, focus on either per-class distribution global provides two need be considered...
Abstract Since the preparation of labeled data for training semantic segmentation networks point clouds is a time-consuming process, weakly supervised approaches have been introduced to learn from only small fraction data. These methods are typically based on learning with contrastive losses while automatically deriving per-point pseudo-labels sparse set user-annotated labels. In this paper, our key observation that selection which samples annotate as important how these used training. Thus,...
We propose a new method for synthesizing an arbitrarily sized novel vector texture given single raster exemplar. Our first segments the exemplar to extract primary textons, and then clusters them based on visual similarity. compute descriptor capture each texton's neighborhood which contains inter-category relationships that are used at synthesis time. Next, we use simple procedure both place secondary textons behind polygons. Finally, our constructs gradient field background is defined by...