Hello!
I am Rajat Aggarwal
B.Tech(Hons) & MS by Research, Computer Science & Engineering

Home

Something about me

Hi! I am Rajat Aggarwal

I am a senior year undergraduate student at International Institute of Information Technology, Hyderabad (Anticipated graduation- 2016). My research interests are Computational Photography, Computer Vision, Human Computer Interaction and Machine Learning. I am doing my research under the guidance of Dr. Anoop M. Namboodiri at Center for Visual Information Technology.

My current research is targeted at Panoramic Stereo Imaging, Virtual Reality, Light Field Representations and Estimation of physical properties(shape, depth, motion, colors, light, reflections etc.) of an object from minimal images.

Services
Home

Publications

    • Rajat Aggarwal, Amrisha Vohra and Anoop Namboodiri, "Panoramic Stereo Videos Using a Single Camera" , CVPR 2016.

    • Rajat Aggarwal, Sirnam Swetha, Anoop M. Namboodiri, Jayanthi Sivaswamy and C.V.Jawahar, "Online Handwriting Recognition Using Depth Sensors" , ICDAR 2015.
  • Research

    Camera-projector sensor for segmentation of specular regions

    Dr. Anoop Namboodiri

    This work is based on computer vision problem of segmenting specular regions in a scene. We propose a camera- projector sensor which automatically segments specular regions with projective information into the world. It involves an expectation-maximization solution to the problem using state-of-the-art segmentation techniques.


    Day-2-Day Day-2-Day Day-2-Day
    Camera-Projector Sensor Design (L), The Projection into the world (C) and Reflection of the world(R).


    Omnistereo Video Camera

    Dr. Anoop Namboodiri

    This work is based on designing a stereo camera which can capture an accurate panoramic videos with dynamic scenes. This camera will solve the inaccuracies in alignment, motion parallax, seams and disparity which come by capturing the scene using hand-held cameras. More details of this work will be updated soon.



    Fast Separation of Direct & Global Illumination of Light

    Dr. Anoop Namboodiri

    This work aims at fast separation of direct and global components of a scene measured by a camera-projector system using projective patterns. Direct components helps in finding the nature of material, 3D Reconstruction and post-processing in segmentation problems. State-of-the-art method solves this problem by the images captured by projecting high frequency binary illumination pattern. We use properties of specular surfaces to identify and separate out the total direct reflections on the specular surfaces.




    Day-2-Day Day-2-Day Day-2-Day
    Original Scene with high frequency binary illumination pattern (L), Direct Component (C) and Global Component (R).

    Real Time Air based handwriting recognition using depth sensors

    Dr. Anoop Namboodiri, Dr. C.V. Jawahar

    The work proposes n-HCI evolution which involves the natural way of handwriting by hand in the air. With most of the depth sensors such as Kinect and Leap Motion, people can actually communicate with the machine by their air handwriting. These systems have been proved to be useful for teaching purposes, blind-deaf communication and it introduces sixth sense technologies in most of the commonly used devices.




    Discovering Recurring Patterns in a single Image

    The project aims at discovering recurring patterns from a single view using the formulation and validation of a joint assignment optimization problem where multiple visual words and object instances of a potential recurring pattern are considered simultaneously. We use two methods with and without pre-knowledge of objects. First approach is achieved by segmentation techniques whereas the optimization is achieved by a greedy randomized adaptive search procedure (GRASP) with moves specifically designed for fast convergence. We have quantified systematically the performance of our approach under stressed conditions of the input (missing features, geometric distortions). We demonstrate that our proposed algorithm outperforms state of the art methods for recurring pattern discovery on a diverse set of real images.


    [ pdf ]

    Day-2-Day Day-2-Day Day-2-Day
    Images adapted from http://vision.cse.psu.edu.
    Day-2-Day

    Wavelet based feature extraction in Pattern Recognition tasks

    Dr. Jayanthi Sivaswamy (CVIT)

    This work aims at the assessment of process parameters or states in a given application using the features extracted from the wavelet coefficients of measured process signals. It is tested on the dataset of 400 textured images of woods from their wavelet coefficients. Learning for the recognition tasks is done on probabilistic neural networks which reported an accuracy of 94%.


    [ pdf ]

    Intelligent Video Surveillance In Automatic Teller Machine

    Dr. Kishore Prahallad (LTRC)

    In this project, we present a novel and efficient fraud detection system where we detect face using color features, extract facial features based on generic features such as edges and structural matching, and determine the presence of fraud based on extracted features.


    [ pdf ]

    Technical Projects

    Comprehensive Shop Billing system SAAS application using Cloud Platform

    Dr. Ganesh Iyer(SDE, Progress Softwares)

    A SAAS Shop billing application deployed on AWS which provides the system on demand service to both users, customers and the retailers. It is a handy application which provides retails choice of interface and billing needs. This application is the part of Progress Rollbase Demo Applications. This application is targeted to the users which are less comfortable with computers and involves less mathematics. Progress Rollbase is a productivity PaaS to rapidly develop and deploy SaaS applications. The application will generate a bill or invoice against the purchase for customers, it will maintain the stock details of different products. This application stores only least require details about the products that are required for processing the bill, like Product name, Product ID, Quantity and Price per unit. The administrator can add different products that he sell in his organization, and the price per unit of that product, he can update the stock details and also he can add or delete customers from the application database. The application will maintain stock details for various products. The operator using the developed application will generate bill for different items that the customer purchase from the shop and will generate a bill/invoice for the same.


    [ Application ]

    Day-2-Day Day-2-Day Day-2-Day

    Online Notater

    Dr. Vikram Pudi (CDE, COS)

    A web based application on web2py framework ,designed for note taking and archiving. A note can be a piece of formatted text, a full webpage or webpage excerpt, a photograph, a voice memo, or a handwritten "ink" note. Notes can also have file attachments. Notes can be sorted into folders, then tagged, annotated, edited, given comments, searched, and exported as part of a notebook. It also allow users to categorize notes using folders or labels, set the reminders, deadlines for their work and notes are searchable based on their tags, whether they have a reminder, audio, images, or are a list. This application is similar to Evernote and Google Keep.


    [ Application ]

    Digital Logic Design Virtual Lab - Android version

    Dr. P.J Narayanan (CVIT, Director-IIITH)

    An Android app that allow users to design and simulate any digital combination or sequential circuit. It allow the users to play with the virtual circuit and outputs the desired results whilst the circuit is running. This application is the Android version of Digital Logic Design Virtual Lab, IIITH.


    [ Application (.apk) ]

    dld dld dld

    3D Bike Game

    Dr. P.J. Narayanan

    A 3D game developed on OpenGL. A synthetic 3D world is developed with the lighting and illumination effects to simulate the real experience.


    Interval Trees & Segment Trees based Query engine

    Dr. Kishore Kothapalli (CSTAR)

    Implementation of Interval Trees and Segment trees in C language using Red -Black trees with efficient algorithms and run-time best in theory. It was tested by large set of queries e.g. Adding two intervals, Searching, Updating etc.


    [ pdf ]

    Linux Shell

    Dr. P.K. Reddy (CDE)

    Implemented a customized Linux shell in C. Basic Linux scripting has been tested on the application. Code can be used as the snippet for building much more customized Linux shell.


    [ Application ]
    dld

    Wave Mixer

    Dr. Kishore Prahallad (LTRC)

    This is the basic application which mixes two sound waves. It also allows users to amplify, modulate, scale and shift the waves. Users can select three audio files at maximum to process through GUI.


    [ Application ]

    NLP based Q/A engine

    Dr. C.V. Jawahar (CVIT)

    Project involves parsing the commentatry given on ESPN CricInfo . This application is build on python using NLTK library. Using NLTK models, this application generalizes the predicates based on the given context-free grammar. Based on the modelling, this application tries to answer the input questions. Some of the example questions are:

    In the first match, Kohli hit sixes in overs of which bowler(s)?
    For all innings, if strike rate of player is above 200.0 then he has hit more sixes than fours? (True/False)?

    Backgammon bot

    Dr. C.V. Jawahar (CVIT)

    Project involves developing Backgammon computer bot based on AI techniques. Strategy uses 1-ply deep expectimax and evaluatio function for optimized move.


    [ Application ]



    Contact

    Contact Information

    phone icon +91 8297001120
    address icon OBH- 220
    IIIT-H ,Gachibowli,
    Hyderabad-500032, Telangana, India
    Home