No talks scheduled yet.
Abstract
Large collections of 3D models are now commonly available via many
public repositories, opening new possibilities for data mining,
visualization, and synthesis of new models. However, exploring such
collections remains challenging because similarity relationships between
points on 3D surfaces are often ambiguous and/or difficult to infer
automatically. To address this challenge, we introduce an encoding of
similarity relationships using fuzzy point correspondences. Based on the
observation that correspondence space is low-dimensional, we propose a
robust and efficient computational framework to estimate fuzzy
correspondences using only a sparse set of pairwise model alignments. We
evaluate our algorithm on a range of correspondence benchmarks and
report substantial improvements both in terms of accuracy and speed
compared to existing alternatives. Further, we use fuzzy correspondences
to process large model collections collectively and demonstrate
applications towards view alignment, smart exploration, and faceted
browsing.
Bio
Janet Vertesi is Link-Cotsen fellow at the Society of Fellows at
Princeton and a Lecturer in Sociology. Her forthcoming book is based
on over two years of ethnographic immersion with the Mars Exploration
Rover mission team. She has worked in Human-Computer Interaction for
over seven years, publishing at CHI, CSCW and Ubicomp; worked with
Phoebe Sengers' Culturally Embedded Computing research group at
Cornell's Information Science department and joined Paul Dourish's
group as a postdoctoral researcher at the Informatics Department at
the University of California Irvine. She is currently PI of a
SocioComputational Systems grant from the NSF to continue her work
with NASA spacecraft teams.
http://janet.vertesi.com
Abstract
Jeff Snyder is interested in the creation of new instruments for
electronic music performance. Much of his research focuses on more
expressive control of musical parameters through the use of new
hardware interfaces. Other topics he engages with include electronic
control of acoustic sound via vibration transducers and custom-built
physical resonators, and electromechanical production of electronic
sound. Bio
Jeff Snyder (b.1978) is a composer, improviser and instrument-designer
living in Princeton, New Jersey, and active in the New York City area.
He performs on analog modular synthesizer in duos with Sam Pluta and
Eric Wubbels, and also leads a band his electro-country alter ego Owen
Lake.
He is currently the Technical Director of the Princeton University
Electronic Music Studios, and the Associate Director of PLOrk, the
Princeton Laptop Orchestra. In 2011, he received a doctorate with
distinction in Music Composition from Columbia University.
In 2009, Jeff started a small business designing and manufacturing
electronic musical instruments under the name Snyderphonics. The same
year, he co-founded an experimental music record label with Sam Pluta
and David Brynjar-Franzson called Carrier Records.
Abstract
This paper investigates ``Schelling points'' on 3D meshes, feature points selected by people in a pure coordination game due to their salience. To collect data for this investigation, we designed an on-line experiment that asked people to select points on 3D surfaces that they expect will be selected by other people. We then analyzed properties of the selected points, finding that: 1) Schelling point sets are usually highly symmetric, and 2) local curvature properties (e.g., Gauss curvature) are most helpful for identifying obvious Schelling points (tips of protrusions), but 3) global properties (e.g., segment centeredness, proximity to a symmetry axis, etc.) are required to explain more subtle features. Based on these observations, we use regression analysis to combine multiple properties into an analytical model that predicts where Schelling points are likely to be on new meshes. We find that this model benefits from a variety of surface properties, particularly when train!
ing data comes from examples in the same object class.
Abstract
Multispectral imaging is a method to obtain the spectrum of each pixel
in a 2D image. Instead of roughly aggregating light arriving at each
pixel into one channel (e.g. gray-scale images) or three channels (RGB
images), multispectral imaging divides the spectrum of each pixel into
much narrower wavelength bands, obtaining a 3D cube of data indexed by
spatial position and wavelength. Some prior works have shown promising
acquisition quality, but they either require tedious manual work and
very long acquisition time, or need active illumination or even
additional cameras. Our work is a single-shot multispectral imaging
method, which tries to capture a multispectral cube with one camera and
within a single camera shot. A diffraction grating is put between camera
lens and the scene to diffract lights from the scene into one clear
zero-order image, and several dispersed first-order images. Each image
is a projection of the 3D spectral cube on a plane. Based on these
images and projections, and based on a prior assumption that camera
noise is with Poisson distribution, an iterative reconstruction
algorithm is applied to infer the 3D spectral cube. So far, several
synthetic experiments have been done to analyze and alleviate the
sensitivity of the acquisition quality with respect to camera noise. We
will also conduct real experiments in the following days.
Abstract
Recent spatially varying reflectance (svBRDF) printing systems can reproduce an input document as a combi- nation of matte, glossy and metallic inks. Due to the limited number of inks, this reproduction process incurs some distortion. In this work, we present an svBRDF gamut mapping algorithm that minimizes distortions in the angular and spatial domains. To preserve a material’s perceived variation with lighting and view, we introduce an improved BRDF similarity metric that builds on both experimental results on reflectance perception and on the statistics of natural lighting environments. Our experiments show better preservation of object color and high- lights, as validated quantitatively as well as through a perceptual study. As for the spatial domain, we show how to adapt traditional color gamut mapping methods to svBRDFs. Our solution takes into account the contrast between regions, achieving better preservation of textures and edges.
Abstract
Image alignment is one of the very first steps for most computer vision algorithms. Image fusion, image mosaicing, image panorama, object recognition and detection, photometric stereo enhanced rendering are some of the examples in which image alignment is a crucial step for a promising result. In this work, we focused on a specific problem which is alignment of high-resolution images taken from the same viewpoint under different light directions. Although images are taken from the same viewpoint, there might be some misalignment due to perturbations to the camera and the effect of optical image stabilization. For this specific alignment problem, we made a broad literature survey and try different tools and algorithms to compare different approaches and find out the best approach to solve this problem. Based on our experiments, we saw that the best approaches are the feature-based ones. We found SIFT and SURF key points reliable for most cases. Datasets that we are interested in consist of many images, and we built minimum spanning tree for each dataset in order to leverage having many images. For feature-based approaches, one of the main problems is elimination of outliers, and we solved this problem with using RANSAC framework. Also, we did some experiments on aligning infrared images to visible images, although other algorithms in the literature do not work well on this problem, our approach gives promising results.
Abstract
Facial caricature is a powerful art form that expresses the individuality of a person by emphasizing the distinctive facial features in an artistic and humorous way. Computer generation of facial caricature from still images have been well-studied in the recent years. Existing work has analyzed the way artists produce caricatures and designed varieties of algorithms that tries to automate this process. However, facial caricature in the video domain is not well explored. Videos different from images pose additional challenges such as feature tracking and temporal coherence. Video as a more expressive media also opens up different ways of making caricatures. Not only the facial features such as eyes, nose etc can be statically enlarged or shrinked, but also the dynamic process of smile and other expressions can be emphasized or subdued, which might result in different perception from the viewers. We developed a simple prototype for facial expression exaggeration on the fly. We are working on improving the caricature quality and designing video stylization techniques that target specifically towards video. As evaluation, we are planning to conduct psychological experiments to explore how expression caricature can be applied to enhance visual communications.
Abstract
Fabrics have a wide range of appearance determined by their small-scale 3D structure. Accurately modeling this structural detail can produce highly realistic renderings of fabrics, and is critical
for predictive rendering of fabric appearance. But building these yarn-level volumetric models is challenging. Procedural techniques are manually intensive, and fail to capture the naturally arising
irregularities which contribute significantly to the overall appearance of cloth.
In this talk I will present a new approach to creating volumetric models of woven fabrics using micro computed tomography (CT) scans. Our approach takes user-specified fabric designs with CT scans of a
small set of cloth samples and produces models that correctly capture the yarn-level structural details of cloth. Bio
Shuang Zhao is a Ph.D. candidate in the Computer Science Department of Cornell University. Supervised by Prof. Kavita Bala, he mainly works on physically-based rendering and appearance modeling.
Previously, Shuang obtained his Bachelor of Engineering degree from the Computer Science & Engineering Department of Shanghai Jiao Tong University.
Shuang visited the Computer Graphics Group at CSAIL of Massachusetts Institute of Technology during academic year 2010-2011.
Abstract
In this project, we propose an automatic algorithm for finding a correspondence map between two 3D surfaces. The key insight is that global reflective symmetry axes are stable, recognizable, semantic features of most real-world surfaces. Thus, it is possible to find a useful map between two surfaces by first extracting symmetry axis curves, aligning the extracted curves, and then extrapolating correspondences found on the curves to both surfaces. The main advantages of this approach are efficiency and robustness: the difficult problem of finding a surface map is reduced to three significantly easier problems: symmetry detection, curve alignment, and correspondence extrapolation, each of which has a robust, polynomial-time solution (e.g. optimal alignment of 1D curves is possible with dynamic programming). We investigate of this approach on a wide range of examples, including both intrinsically symmetric surfaces and polygon soups, and find that it is superior to previous methods in cases where two surfaces have different overall shapes but similar reflective symmetry axes, a common case in computer graphics.
Abstract
Energy efficiency is one of the most important topics of the next
years and plays an important role in architecture and building design.
The use of sunlight to illuminate the interior of buildings
(daylighting) contributes to this to a large degree. In this project I
will investigate the possibilities of how to optimize window blinds
with techniques normally used in Computer Graphics. The optimized
window blinds will be used to control the electromagnetic energy
(especially visibible light) flow into buildings. In order to do this,
we are trying to optimize the micro geometry and structure of
reflecting blinds as well as the way they are arranged. This is work
in progress, so your suggestions will help direct the project.
|
