Computer Vision Engineer

Vishnukumar Galigekere N

Phone: 937-***-****

Email: *************@*****.***

Address: **** **** **. *** ****, Irving TX 75039

LinkedIn: http://www.linkedin.com/profile/view?id=17770748&trk=nav_responsive_tab_profile

OBJECTIVE

Seeking a full time position as a software engineer/developer in a challenging and research oriented environment.

My areas of interest are Computer Vision and Machine Learning.

TECHNICAL SKILLS

Proficient in: C, C++, MATLAB

Familiar with: Java, Perl, R, SQL, Python, HTML

Platforms: Windows, Linux

EDUCATION

Ph.D. in Computer Science (2013)

University of Texas at Arlington, Arlington, TX

Advisor: Dr. Gutemberg Guerra-Filho

Thesis Title: An Interpolation Based Approach for Pattern Recognition and Generation

M.S. in Computer Science (2008)

University of Texas at Arlington, Arlington, TX

Advisor: Dr. Nikola Stojanovic

Thesis Title: Computational Analysis of Variability of Recombination Rates in Mice

Master of Computer Applications (2004)

Visveswaraya Technological University, Bangalore, India

RESEARCH PROJECTS

Object/Face Recognition in Varying Illumination Conditions

Designed a novel interpolation based supervised learning algorithm to model lighting variation in images. The

model allows for both synthesis and analysis of images in different illumination conditions such as lighting

intensity and/or lighting position

Devised and successfully implemented an image recognition framework over the interpolation algorithm to

consider varying illumination conditions

Designed and captured an image dataset with varying lighting intensity to demonstrate and publish experimental

results

Evaluated the novel recognition method with our image dataset and the Yale extended face database for varying

lighting position and showed extremely competitive results in terms of recognition rate

Image Based Rendering and Lighting

Designed a novel Image Based Rendering and Lighting algorithm using a synthesis-analysis technique along

with a novel interpolation based supervised learning algorithm

The Image Based Rendering and Lighting algorithm can render nonexistent objects in unknown environments

as long as the object models are learnt using the interpolation based learning algorithm

The same algorithm is also capable of inferring the lighting parameters of unknown environments provided an

object whose interpolation model is already learnt exists in the scene

Motion Retargeting and Analysis

A novel interpolation based supervised learning algorithm is used to mathematically model animated motion

using a set of motion capture data as training data

The model represents a particular action for a set of all possible (topologically identical) skeletons/characters

defined within a parametric space

With this model for a particular trained motion sequence the same motion sequence is retargeted into a new

character with different skeletal parameters

The inverse computation of this interpolation model is used to recognize the skeleton’s identity given its motion

sequence

Topological Gabor Descriptor

Designed and constructed an image descriptor invariant to scale, rotation about the Z axis and uniform lighting

variations

The filter bank is built using a novel arrangement of 2D Gabor filters. The topology of the filter bank allows for

a simple 1D circular shift in the descriptor space to search for matches that are rotationally invariant. The

descriptor is essentially like a time series comprising of the responses of the filters in the filter bank. Image

matching or recognition is achieved by performing a circular shift in the descriptor space while computing and

minimizing a suitable distance metric

The descriptors are computed for a spatial location in a dense pyramid of images to achieve scale invariance.

The rotational invariance is robust to a complete 360 degree rotation in the z-axis

Estimating Recombination Variability among Mouse Strains

Developed scripts to extract large amounts of genetic data from web pages and compiled them into data-sets

Constructed genetic-maps for each data-set utilizing QTL (Quantitative Trait Loci) mapping tools using Hidden

Markov Models (HMM)

Statistically analyzed the resulting genetic-maps to study the genetic control over the rate of meiotic

recombination in mice

RELEVANT ACADEMIC PROJECTS

AdaBoost Based Face Recognition System

Implemented an AdaBoost based face detection system using Haar-like features (Rectangle filters) as weak

classifiers on integral-images

Implemented a face recognition system using ADABoost and one-vs-all classifiers

A Principal Component Analysis (PCA) Based Face Recognition System

Implemented a face recognition algorithm using Principal Component Analysis and Linear Discriminant

Analysis and tested with the ORL face database

Stacking Blocks using RHINO Robot Arm

A robot system consisting of a 6 degree of freedom robot arm and a camera was programmed to pick up blocks

off the table and build a particular set of two stacks from them (without knocking off any stacks). The

implementation was divided into the following subtasks;

Performed image processing to compute the pixel coordinates of the object

Converted pixel coordinates into world coordinates

Performed inverse kinematics and picking and stacking of blocks in the stacking area

Navigating a Known Obstacle Course with a Lego Robot

Designed and built a holonomic robot to navigate through an obstacle course (obstacles, start and goal locations

are selected arbitrarily)

Implemented a navigation strategy using Manhattan distances to generate all possible paths from Start to Goal

The robot iteratively computes the values for each neighboring cells and uses a greedy approach to reach the

goal with minimum collisions

Behavior-based Fire Alarm Robot

Designed and built a behavior-based fire detection robot that is able to move from an unknown position in an

indoor environment to look for a fire and raise an alarm

A set of three behaviors, including "wander" (search), "wall following" and "fire detection"(and raise alarm)

were implemented

These behaviors were integrated using the subsumption architecture

PUBLICATIONS

[1] Vishnukumar Galigekere N, Gutemberg Guerra-Filho, “An Interpolation Based Approach for Motion

Retargeting and Recognition” (currently in progress)

[2] Vishnukumar Galigekere N, Gutemberg Guerra-Filho, “A Synthesis-and-Analysis Approach to Image Based

Lighting”, International Symposium on Visual Computing (ISVC) 2012.

[3] Vishnukumar Galigekere N, Gutemberg Guerra-Filho, “An Interpolation Based Approach for Lighting Variation

in Image Recognition”, IEEE International Conference on Image Processing (ICIP) 2012.

[4] Vishnukumar Galigekere N, Gutemberg Guerra-Filho, “Topological Gabor Descriptors: Exploring a Filter Bank

Structure for Image Feature Matching”, PErvasive Technologies Related to Assistive Environments (PETRA) 2011.

DATASETS AND TOOLS

CVPoV: An Automated Tool for Generating Synthetic Ground Truth Images

CVPoV is a computational tool built over a popular ray-tracer POVRay that allows the user to specifically

generate images/videos of simple objects or complex scenes involving several different objects.

The tool saves depth information and camera parameters which are then used to compute point correspondence,

motion field data, and occlusion maps.

This tool is designed to allow the user to automatically generate synthetic ground-truth data for several

computer vision problems such as camera calibration, feature matching, 3D reconstruction and object

recognition

Image Dataset with Varying Lighting Intensity

Designed and built an image capture setup with 16 identical cameras (Point Grey Flea3) and a static scene under

81 different illumination conditions

Four identical lamps with three intensity levels each were used to create the 81 different lighting configurations

Image Dataset with Varying Geometric Pose

Designed and captured images of several individual objects with 360 rotation about the vertical axis and

translation in x and y axes with a black background

RELEVANT GRADUATE LEVEL COURSEWORK

Digital Image Processing, Computer Vision, Machine Learning, Pattern Recognition, Artificial Intelligence,

Autonomous Robots, Bioinformatics Algorithms, Engineering Genetics, Data Modeling and Analysis

WORK EXPERIENCE

Graduate Teaching Assistant at the Department of Computer Science and Engineering, University of Texas

at Arlington, TX (2007 to 2013)

Assisted the instructor for the courses mentioned below. My role included teaching lab exercises or regular classes

and assisting students with coursework and grading exams and projects.

Intermediate Programming in C

Special Topics in Multimedia, Gaming and Animation

Pattern Recognition

Bioinformatics Algorithms

Software Analyst at SEIMENS, Bangalore, India (Mar 2005 to Jul 2005)

Involved in Testing of Business Processing (Web) application EnCoRRe on Global Distribution Systems (GDS)

like Sabre, Worldspan, and Apollo for various Travel Clients like Worldtravel, Expedia, American Express, and

Boeing

Analyzing client requirements from the Functional Requirements Document

Development of business components for automated transaction processing/data processing

Implementation, testing and integration of individual modules

Customization of the application for various clients

Customization and/or implementation of rules for data processing and reporting. Involves use of powerful

Oracle databases for data storage and security

Internship at Aeronautical Development Agency, Bangalore, India (Mar 2004 to Aug 2004)

A Direct Drive Valve flight control actuator, quadruplex in electrical redundancy and dual hydraulic

redundancy was modeled using the Simulink toolbox

Implemented a Redundancy Management scheme to detect failure, isolate the failed component/s and

automatically compensate for current(motor)/pressure(hydraulic) with the remaining functional components

making it a Fault Tolerant System

ACADEMIC AWARDS AND HONORS

STEM Tuition Fellowship

Graduate Teaching Assistantship

Graduate Research Assistantship

UTA College of Engineering Travel Support for ICIP 2012

Contact this candidate