Ioannis Gkioulekas

I am a PhD candidate in the School of Engineering and Applied Sciences at Harvard University. My advisor is Todd Zickler.

Before coming to Harvard, I received a Diploma (five-year degree) in Electrical and Computer Engineering from the National Technical University of Athens, Greece, where I worked with Petros Maragos.

email | cv | google scholar

that's me
Research Interests

I am interested in computer vision, computer graphics, and machine learning. I am particularly interested in various aspects of material appearance: how humans perceive real world materials, how we can measure their parameters, and how to reason about them in images of the world.


Micron-scale Light Transport Decomposition Using Interferometry
Ioannis Gkioulekas, Anat Levin, Frédo Durand, Todd Zickler,
ACM Transactions on Graphics (SIGGRAPH), 2015
paper | project page

We present a computational imaging system that produces light-transport decompositions at very high spatial and pathlength resolution. It can be used to measure 3D shape, direct and indirect illumination components, and scattering, dispersive and birefringent materials.


On the Appearance of Translucent Edges
Ioannis Gkioulekas, Bruce Walter, Edward H. Adelson, Kavita Bala, Todd Zickler
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015
paper | project page

We study the appearance of edges created by geometric discontinuities in translucent objects. Through physics and simulations, we show that edge profiles exhibit a large variety of features that provide rich information about the underlying material and geometry.


Looking Against the Light: How Perception of Translucency Depends on Lighting Direction
Bei Xiao, Bruce Walter, Ioannis Gkioulekas, Todd Zickler, Edward H. Adelson, Kavita Bala
Journal of Vision (JOV), 2014
paper | project page

We use appearance-matching perceptual experiments to study the effect of lighting, geometry and material parameters on how humans perceive translucent objects.


Inverse Volume Rendering with Material Dictionaries
Ioannis Gkioulekas, Shuang Zhao, Kavita Bala, Todd Zickler, Anat Levin
ACM Transactions on Graphics (SIGGRAPH Asia), 2013
paper | project page

We present a system for measuring the bulk scattering properties of homogeneous material samples. The proposed system can be used for both solid and liquid materials of various scattering densities, and does not restrict the shape of the phase function.


Understanding the Role of Phase Function in Translucent Appearance
Ioannis Gkioulekas, Bei Xiao, Shuang Zhao, Edward H. Adelson, Todd Zickler, Kavita Bala
ACM Transactions on Graphics 2013 (Presented at SIGGRAPH 2013)
paper | project page

We generalize scattering phase function models, demonstrate an expanded translucent appearance space, and discover perceptually-meaningful translucency controls by analyzing thousands of images with computation and psychophysics.


Effects of Shape and Color on the Perception of Translucency
Bei Xiao, Ioannis Gkioulekas, Asher Dunn, Shuang Zhao, Todd Zickler, Edward H. Adelson, Kavita Bala
Vision Science Society Annual Meeting (VSS), 2012
abstract | slides | data

We study how translucency perception is related to color and object geometry, by measuring perceptual similarities between images rendered with varying combinations of scattering parameters, color and shape.


Toward Wide-Angle Microvision Sensors
Sanjeev J. Koppal, Ioannis Gkioulekas, Travis Young, Hyunsung Park, Kenneth B. Crozier, Geoffrey L. Barrows, Todd Zickler
IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 2013
paper | project page

This is the journal version of our CVPR 2011 paper and includes more detailed analysis, new miniaturized prototypes, and a deployment on a micro-robotic air vehicle.


Dimensionality Reduction Using the Sparse Linear Model
Ioannis Gkioulekas, Todd Zickler
Advances in Neural Information Processing Systems (NIPS), 2011
paper | project page

We use the sparse linear model to learn low-dimensional linear projections that approximately preserve inner products between sparse codes, used to represent various classes of signals.


Wide-angle Micro Sensors for Vision on a Tight Budget
Sanjeev J. Koppal, Ioannis Gkioulekas, Todd Zickler, Geoffrey L. Barrows
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2011
paper | project page

We propose and analyze a class of vision sensors that can be used on micro-scale devices with severe power and mass constraints, for tasks such as locating edges, tracking targets, and detecting faces.


Spatial Bayesian Surprise for Image Saliency and Quality Assessment,
Ioannis Gkioulekas, Giorgos Evangelopoulos, Petros Maragos
IEEE International Conference on Image Processing (ICIP), 2010
project page

We use an information-theoretic approach to study bottom-up spatial saliency, and apply our model for attention modeling and fixation prediction, image region detection, and image quality assesement.


CS283 Computer Vision - Fall 2015 (Harvard)

CS283 Computer Vision - Fall 2014 (Harvard)

CS283 Computer Vision - Fall 2013 (Harvard)

CS283 Computer Vision - Fall 2012 (Harvard)

CS283 Computer Vision - Fall 2010 (Harvard)


All of my released code is hosted on my GitHub account. I use the coding and style conventions described here in all of my projects.

libslm A machine learning library with emphasis on regularized (sparsity, nuclear norm, group norms) optimization and kernelization, for problems such as regression, coding, dimensionality reduction, and metric learning. Written in C/C++, with MATLAB wrappers provided.
blas_ext A linear algebra library of extensions to BLAS and LAPACK for dense matrices. Includes Cholesky and QR updates, downdates and row/column modifications, SVD and EIG updates, ports of PROPACK and ARPACK, Nystrom methods, and other randomized decompositions. Written in C, with MATLAB wrappers provided.
image_utils A library of image processing utilities. Includes implementations of many algorights for high dynamic range imaging, color transformations and editing, hyperspectral image editing, as well as wrappers for various image formats (HDR and RAW). Written in C/C++ and MATLAB.
mex_utils A set of C++ header files for interfacing with the MATLAB mex API.
Last updated September 2015. Valid XHTML 1.0 Transitional Valid CSS!