Cheuk Yiu Ip ( I have joined AT&T Labs Research )

Visualization and Computer Graphics

PixelPie: Maximal Poisson-disk Sampling with Rasterization

We present PixelPie, a highly parallel geometric formulation of the Poisson-disk sampling problem on the graphics pipeline. Traditionally, generating a distribution by throwing darts and removing conflicts has been viewed as an inherently sequential process. In this paper, we present an efficient Poisson-disk sampling algorithm that uses rasterization in a highly parallel manner. Our technique is an iterative two step process. The first step of each iteration involves rasterization of random darts at varying depths. The second step involves culling conflicted darts. Successive iterations identify and fill in the empty regions to obtain maximal distributions. Our approach maps well to the parallel and optimized graphics functions on the GPU and can be easily extended to perform importance sampling. Our implementation can generate Poisson-disk samples at the rate of nearly 7 million samples per second on a GeForce GTX 580 and is significantly faster than the state-of-the-art maximal Poisson-disk sampling techniques.

Parallel Geometric Classification of Stem Cells by their Three-Dimensional Morphology

Recent work indicates that the physical structure of a tissue engineering scaffold can direct stem cell function by driving stem cells into morphologies that induce their differentiation. Thus, we have developed a rapid method for classifying cells based on their 3D shape. First, random lines are intersected with 3D Z-stacks of confocal images of stem cells. The intersection lengths are stored in histograms, which are then used to train a support vector machine (SVM) learning algorithm to distinguish between stem cells cultured on differentiation-inducing 3D scaffolds and those cultured on non-differentiating flat substrates. Our algorithm is accelerated by a parallel GPU implementation and the trained SVM is able to properly classify the 'new' query cells over 80% of the time.

Hierarchical Exploration of Volumes Using Multilevel Segmentation of the Intensity-Gradient Histograms

Visual exploration of volumetric datasets to discover the embedded features and spatial structures is a challenging and tedious task. In this paper we present a semi-automatic approach to this problem that works by visually segmenting the intensity-gradient 2D histogram of a volumetric dataset into an exploration hierarchy. Our approach mimics user exploration behavior by analyzing the histogram with the normalized-cut multilevel segmentation technique. Unlike previous work in this area, our technique segments the histogram into a reasonable set of intuitive components that are mutually exclusive and collectively exhaustive. We use information-theoretic measures of the volumetric data segments to guide the exploration. This provides a data-driven coarse-to-fine hierarchy for a user to interactively navigate the volume in a meaningful manner.

Saliency-Assisted Navigation of Very Large Landscape Images

This work presents navigation of very large landscape images from an interactive visualization perspective. The grand challenge in navigation of very large images is identifying regions of potential interest. We show that our approach of progressive elicitation is fast and allows rapid identification of regions of interest. We validate the results of our approach by comparing them to Internet user-tagged regions of interest on several very large landscape images.

MDMap : A System for Data-Driven Layout and Exploration of Molecular Dynamics Simulations

MDMap is an automated system to visualize MD simulations as state-transition diagrams, and can replace the current tedious manual layouts of biomolecular folding landscapes with an automated tool. The layout of the representative states and the corresponding transitions among them is presented to the user as a visual synopsis of the long-running MD simulation.

Social Snapshot: A system for temporally coupled social photography

Social Snapshot actively acquires and reconstructs temporally dynamic data. The system enables spatiotemporal 3D photography using commodity devices, assisted by their auxiliary sensors and network functionality. It engages users, making them active rather than passive participants in data acquisition.

Salient Frame Detection for Molecular Dynamics Simulations

Saliency-based analysis can be applied to time-varying 3D datasets for the purpose of summarization, abstraction, and motion analysis. As the sizes of time-varying datasets continue to grow, it becomes more and more difficult to comprehend vast amounts of data and information in a short period of time. In this paper, we use eigenanalysis to generate orthogonal basis functions over sliding windows to characterize regions of unusual deviations and significant trends. Our results show that motion subspaces provide an effective technique for summarization of large molecular dynamics trajectories.

Saliency Guided Summarization of Molecular Dynamics Simulations

We present a novel method to measure saliency in molecular dynamics simulation data. This saliency measure is based on a multiscale center-surround mechanism, which is fast and efficient to compute. We explore the use of the saliency function to guide the selection of representative and anomalous timesteps for summarization of simulations.

Computer Integrated Manufacturing

Retrieving Matching CAD Models by Using Partial 3D Point Clouds

The ability to search for a CAD model that represents a specific physical part is a useful capability that can be used in many different applications. This paper presents an approach to use partial 3D point cloud of an artifact for retrieving the CAD model of the artifact. We assume that the information about the physical parts will be captured by a single 3D scan that produces dense point clouds. CAD models in our approach are represented as polygonal meshes. Our approach involves segmenting the point cloud and CAD mesh models into surface patches. The next step is to identify corresponding surface patches in point clouds and CAD models that could potentially match. Finally, we compute transformations to align the point cloud to the CAD model and compute distance between them. We also present experimental results to show that our approach can be used to retrieve CAD models of mechanical parts.

Computer-Aided Design and Solid Model Matching

A 3D Object Classifier for Discriminating Manufacturing Processes

This work addresses the problem of manufacturing process discrimination, i.e.,determination of the best (or most likely) manufacturing process from shape feature information. We introduce a new shape descriptor with the statistics of surface curvatures. We use support vector machines to learn a separator to classify models that are "prismatic machined" and "cast-then-machined".

Benchmarking CAD Search Techniques

This work presents several distinctive benchmark datasets for evaluating techniques for automated classification and retrieval of CAD objects. These datasets include (1) a dataset of CAD primitives (such as those common in constructive solid geometry modeling); (2) two datasets consisting of classes generated by minor topological variation; (3) two datasets of industrial CAD models classified based on object function and manufacturing process, respectively; (4) and a dataset of LEGO models from the Mindstorms robotics kits.

Automated Learning of Model Classifications

This work describes a new approach to automate the classification of solid models using machine learning techniques. We instroduce a shape learning algorithm and a general technique for "teaching" the algorithm to identify new or hidden classifications that are relevant in many engineering applications.

Using Shape Distributions to Compare Solid Models

This work examines how to adapt shape distributions techniques to comparison of 3D solid models. First, we show how to extend basic distribution-based techniques to handle CAD data in mesh representation. These extensions address specific geometries that occur in mechanical CAD data. Second, we describe how to use shape distributions to directly interrogate solid models. Lastly, we show how these techniques can be put together to provide a "query by example" interface to a large, heterogeneous, CAD database.