Sheetal Lahabar

From Msplacements

Url: http://students.iiit.ac.in/~sheetal
E-mail: sheetal@research.iiit.ac.in
Contact No: +91 9866217899

Career objective
To obtain a position that will enable me to use my strong programming and organizational skills and ability to work well with people.

Education

  • B. Tech + M.S. by research (CSE) , IIIT-Hyderabad, Sept, 2009 (expected); CGPA: 9.03
  • Senior Secondary, Shivaji Science College, Nagpur, Maharashtra, 2003, 88.5%
  • Secondary, Mary Immaculate Girls High School, Mumbai, Maharashtra, 2001, 83.6%

Publications

  • Sheetal Lahabar and P. J. Narayanan, “Singular Value Decomposition on GPU using CUDA”, 23rd IEEE International Parallel and Distributed Processing Symposium (IPDPS 2009), Rome, Italy, May 2009.
  • Sheetal Lahabar, Pinky Agrawal and P. J. Narayanan, “Computer Vision on GPU”, National Academy of Sciences Journal (NAS 2008), Allahabad, India.
  • Sheetal Lahabar, Pinky Agrawal and P. J. Narayanan, “High Performance Pattern Recognition on GPU”, 1st IEEE National Conference on Computer Vision, Pattern Recognition, Image Processing and Graphics (NCVPRIPG 2008), Ahmedabad, India, Jan. 2008.

Academic Achievements

  • Have been in the Dean List for Academic Excellence in Monsoon 2004, Monsoon 2005, Spring 2005, Monsoon 2006, Spring 2006 and Spring 2007.
  • Secured an All India Rank 634 in AIEEE 2004, State rank 19 in Maharashtra.
  • Received the 23rd International Parallel and Distributed Processing Symposium 2009 student travel grant.
  • Was awarded a scholarship by IARCS to attend IPDPS 2009 held in Rome, Italy.

Professional Experience

  • Working as a paid research assistant(July 2008-till date)
    Center for Visual Information Technology Lab, IIIT Hyderabad, Hyderabad
    Advisor: Prof. P. J. Narayanan. Email: pjn@iiit.ac.in
  • Summer Internship as an honors student(May-July 2006 and 2007)
    Center for Visual Information Technology Lab, IIIT Hyderabad, Hyderabad
    Project : High Performance Pattern Recognition on the GPU
    Advisor: Prof. P. J. Narayanan. Email: pjn@iiit.ac.in
  • Summer Internship(May-July 2005)
    Center for Visual Information Technology Lab, IIIT Hyderabad, Hyderabad
    Project : Image Layout Analysis and Reconstruction in HTML
    Advisor : Prof. C. V. Jawahar. Email: jawahar@iiit.ac.in
  • Teaching Assistant(Monsoon 2007)
    Subject: Spatial Informatics
    Instructor: Dr. K. S. Rajan. Email: rajan@iiit.ac.in
  • Teaching Assistant(Spring 2006)
    Subject: Artificial Intelligence
    Instructor: Dr. Rajeev Sangal. Email: sangal@iiit.ac.in
  • Teaching Assistant(Monsoon 2005)
    Subject: Physics I
    Instructor: Prof. Harjinder Singh. Email: laltu@iiit.ac.in
  • Worked on a semester project in Center for Data Engineering, IIIT Hyderabad(Spring 2006)
  • Project leader for the Video Google project(Spring 2007)

Research Projects

  • Ray tracing parametric patches on GPU(On going)
    Thesis project
    Advisor: Prof. P. J. Narayanan
    We aim at utilizing the GPUs for the purpose of direct ray tracing of B`zier patches. A bezier surface is defined by a set of control points. All the existing algorithms are far from interactive performance and the high complexity of their implementation places more demands on the implementation. Direct ray tracing requires computing the eigen values of many small matrices in parallel. Our goal is to design a parallel and efficient eigen value computation algorithm which maps well to the threaded GPU architecture using CUDA.
  • Singular Value Decomposition on GPU using CUDA
    Thesis project
    Advisor: Prof. P. J. Narayanan
    The project aims towards using the GPU for linear algebra algorithms. We implemented the Singular Value Decomposition(SVD) algorithm on the GPU using CUDA using the twin step of Bidiagonalization followed by the Diagonalization. Bidiagonalization is implemented using a series of Householder transformations which map well to BLAS operations. Diagonalization is performed by a sequence of Givens rotations. It gives a speedup of upto 60 and 8 over the MATLAB and Intel MKL implementation respectively on a Intel Dual Core 2.66GHz PC on NVIDIA GTX 280. We have a publication on this work at IEEE International Parallel and Distributed Processing Symposium ’09 as “Singular Value Decomposition on GPU using CUDA”.
  • High Performance Pattern Recognition on GPU
    Honors project in CVIT, IIIT-H
    Advisor: Prof. P. J. Narayanan
    We developed efficient algorithms on Graphics Processing Units(GPU) for pattern recognition problems that are compute intensive. We specifically studied the Parzen windows scheme for density estimation and the Artificial Neural Network(ANN) scheme for training and classification. Our ANN implementation using CUDA achieved 170-210 times speedup over MATLAB and 230-270 times speedup over Fast Artificial Neural Network library on GTX 280. We have a publication on this work at National Conference on Computer Vision, Pattern Recognition, Image Processing and Graphics ’08 as “High Performance Pattern Recognition on GPU”. Our work was also accepted for inclusion in the National Academy of Sciences Journal.
  • Computer Vision on the GPU
    Honors project in CVIT, IIIT-H
    Advisor: Prof. P. J. Narayanan
    We proposed a design and implementation of a general computer vision toolkit using the CUDA programming model. Our design is based on OpenCV, a popular computer vision toolkit. We maintained similar programmer-level representations for the objects including images and maintain the same calling conventions and formal parameters. We implemented basic image processing techniques (filtering, Canny edge detection), Optical Flow and Tracking and Background subtraction. It provides a speedup of 10-30 on different operations on the NVIDIA GTX 280 GPU.
  • vGoogle: A google approach to video retrieval Course project for Computer Vision
    Advisor: Prof. C. V. Jawahar
    The goal of the Video Google project was to retrieve the key frames and shots of a video containining a particular object or scene with the ease, speed and accuracy with which Google retrieves text documents containing particular words. Preprocessing allows Google to keep track of the ranked list of documents that corresponds to every word which requires a merging of ranked documents for every query term to form the result. We attempted to use this standard approach to recognition. We were able to identify Visual vocabularies on different sets of videos. The results were encouraging and evaluation schemes were used to show the correctness of the implementation.

Course Projects

  • Indic Spell Checker
    Course: Information Technology Workshop
    The aim of the project was to develop a Spell Checker GUI for Indian languages. We used Aspell language dictionaries and developed an interface using Qt for spell checking. It has all the functionalities of a standard spell checker and could be easily extended for other languages.
  • Fractal Image Compression
    Course: Digital Image Processing
    The aim was to implement the fractal image compression to achieve high compression rates, for images which are in general non-fractals using the concepts of Iterative Function System and the Collage Theorem.
  • Hospital Management System
    Course: Introduction to Middleware Systems
    We developed a database system for a Hospital. The development process included problem specification, requirements identification, ER model diagram, Relational model design, populating a database, Web-service design that satisfy the functionality requirement of the user, analyzing the queries, indexing and denormalization to improve the performance.
  • Implementation of the BASH shell with standard functionalities
    Course: Operating Systems
    Simulation of the BASH shell of Linux using system calls with some minor functionalities such as redirection operators, maintenance of history, maintaining background processes etc.
  • Sort Merge Join
    Course: Database Management System
    We implemented the Sort Merge Join used in the implementation of a relational database management system. The aim of the project was to find for each distinct value of the join attribute, the set of tuples in each relation which displays that value. The relations are first sorted by the join attribute so that interleaved scans will encounter these sets at the same time.
  • Building a database in MSSQL
    Course: Database Management Systems
    Given a domain, performing the requirement gathering, building and Entity-Relationship model, creating and maintaining of a database, populating the database with meaningful data and running queries on the data are done as a part of the project using MS Visio and MSSQL.
  • 2D Tetris Game
    Course: Computer Graphics
    To write the tetris game in 2D with functionalities of maintaining score and speed using OpenGL, Glut, SDL etc.
  • 3D simulation of train
    Course: Computer Graphics
    Simulated a 3D train on tracks with functionalities of maintaining the speed of the train, etc.
  • Implementing the standard OpenGL library functions
    Course: Computer Graphics
    Implemented some of the most commonly used OpenGL library functions which include Modeling transformations, Viewing transformations, Projection transformations, Viewport transformations and support for drawing OpenGL primitives.

Courses taken

Computer Graphics, Digital Image Processing, Pattern Recognition, Computer Vision, Signals and Systems, Theory of Computation, Cryptography, Data Compression, Linear Programming, Artificial Intelligence, Introduction to Middleware Systems, Computer Organization, Spatial Informatics, Computer Networks, Compilers, Software Engineering, Principles of Programming Languages, Economics, Database Management Systems.

Academic Interests
General Purpose Computation on Graphics Processing Units, Pattern Recognition, Computer Vision, Image Processing, Data Compression, Cryptography and Database Systems

Technical Skill Set

  • Programming Languages:- C, C++, Java, Lex & Yacc, CUDA, Cg(C for Graphics), J2ME, Oz
  • Operating Platforms:- GNU/Linux, Windows 95/98/2000/XP
  • Scripting Languages:- Perl, Shell, Matlab, Sed, Awk
  • Web Languages:- HTML, XHTML, XML, CGI, PHP, CSS, Apache Tomcat
  • Programming Environments:- GNU/GCC, Microsoft Visual Studio, Vim, Phoenix RDK
  • Libraries:- SDL, Qt, OpenGL, Glut, OpenCV, FANN, FFMPEG
  • Tools:- MySQL/MSSQL, Photoshop/Gimp
  • Publishing:- Latex


Miscellaneous

  • I attended National Conference on Computer Vision, Pattern Recognition, Image Processing and Computer Graphics (NCVPRIPG 2008) at DAIICT Ahmedabad.
  • I attended the International Parallel and Distributed Processing Symposium (IPDPS 2009) at Rome, Italy.
  • Demonstrated the project Smart Phone Book for Mobile Devices at the Research and Development showcase 2006 at IIIT Hyderabad.
  • Demonstrated the Video Google Project at the Research and Development showcase 2007 at IIIT Hyderabad.
  • Completed Elementary level course in German, organized by Goethe-Zentrum, a prestigious German Cultural Center in June 2008.
  • I was one of the organizer of Cultural Events, at Felicity’07, annual technical and cultural festival at IIIT.
  • Winner of inter house throw ball competitions held in 2006 and 2007.
  • Winner of Western Group Dance in Felicity ’07.

Personal Profile

  • Nationality: Indian
  • Languages: English, Hindi and Marathi
  • Passport: Yes
Retrieved from "http://www.iiit.ac.in/msplacements/index.php/Sheetal_Lahabar"
Personal tools