About me

I completed my Master's from the Department of Computer Science at Stony Brook University, NY under the guidance of Prof. Arie Kaufman. My research interests lie at the intersection of Computer Vision and Computer Graphics. I basically love to play with pixels and am interested in solving variety of problems in the domain. I also worked closely with Prof. Joseph Mitchell on solving Art Gallery Problem and its various applications in lighting design, sensor coverage and virtual tours for 3D scenes.

In the past, I received a bachelor's degree from the LNM Institute of Information Technology, Jaipur, India where I spent four wonderful years and made life long friends. During my summer internships, I got golden opportunities to work at CERN (2011), Google Summer of Code'12 and Amazon.com (2013). I learnt a lot through these experiences about different areas of Computer science.

I am from the city of Jaipur, also known as the Pink City of India, which is a world heritage site famous for many historical monuments like Hawa Mahal, Amber Fort, etc.

Stony Brook University, NY (2013-present)
Master's of Science, Computer Science
CGPA - 3.88/4.0

Courses taken:
Fall'13 -- Operating Systems, Analysis of Algorithms, Artificial Intelligence
Spring'14 -- Computational Geometry, Computer Networks, Medical Imaging
Fall'14 -- Computer Graphics, Theory of Database Systems


The LNM Institute of Information Technology, Jaipur, India (2008-2012)
Bachelor of Technology, Electronics and Communication
CGPA - 8.6/10.0

Relevant Courses:
Data Structures, Signal Processing, Digital Communication, Mobile and Broadband Communication, Coding Theory, Computer Vision and Pattern Recognition, Computer Networks, Remote Sensing and GIS.


Amazon.com, Seattle (January, 2015 - present)
Software Developer Engineer, Fulfillment Center Systems

Research and Development work at Amazon Fulfillment Technology Computer Vision Team for developing scalable machine vision software for Amazon warehouses.


Research Assistant, Stony Brook University, NY (August, 2013 - December, 2014)
Center for Visual Computing, Visualization Lab
Advisor: Prof. Arie Kaufman

Virtual Shopping Experience project aimed to create a virtual environment in which a user can try out clothing before making purchases. My project on "3D Full Body Avatar Modelling with Microsoft Kinect" focused on creating a realistic 3D human model to be integrated in the shopping experience. I developed and proposed methods for point cloud modelling using Rigid and Non-rigid registration methods like Iterative Closet Point, Articulation and Features based registration. For gaming application we proposed to estimate human body parameters and fit the parameters in PCA space of a society of human models. (Implemented using Point Cloud Library, C++, Kinect SDK)


Amazon.com, Seattle (May - August, 2014)
Software Developer Intern, Associates Technology

As a part of the Targeting platform team, I added features to target customers for prospective products based on behavioural patterns when they visit Amazon.com.

Winner of Amazon Global Intern Hackathon (450 participants/79 teams) on Computer Vision and Machine Learning organized by Amazon Fulfilment Team.


Google Summer of Code 2012
Open Source Developer
Organization- DIFFER (Determinator of Image File Format PropERties)

Multi-threaded implementation of new similarity metrics (image data) namely, MSE, SSIM, MS-SSIM, PSNR, Universal Quality Index in OpenCV using OpenCL and CUDA support (GPU programming) into existing DIFFER framework. https://bitbucket.org/dushyantgoyal/gsoc-12-differ


European Organization for Nuclear Research(CERN), Switzerland (June - November, 2011)
Research Associate
Advisor: Dr. Federico Carminati, Dr. L. Batev

AliEn is a lightweight Open Source Grid Framework built around other Open Source components using the combination of a Web Service and Distributed Agent Model for simulation, reconstruction, and analysis of physics data of the ALICE experiment.
As a member of ALICE core offline team, I developed a new distributed Grid file catalogue schema which reduced the database size by a factor and increased the access speed by a factor of 5. Report

Fast and efficient Monte-Carlo particle transport code particularly for high energy nuclear and particle physics experiments has become an important tool starting from the design and fabrication of detectors to the modeling of the physics outcome as close as the reality.
We developed such a ROOT based Monte-Carlo application package for low energy neutron transport using the evaluated nuclear data file (ENDF). We proposed a fast and accurate sampling technique for multi dimensional probability distributions to be used in the simulation of these experiments. We also investigated the use of GPUs for leveraging the embarrassingly parallel nature of these computations. .


Information Cryptology and Construction Lab (ICCL), Taiwan (May - August, 2011)
Research Intern
Advisor: Prof. Shiuh-Jeng Wang

Involved in the research on "Reversible Data Hiding" and "Computer Forensics". Developed schemes in "reversible data hiding" for multimedia security.

3D Full Body Avatar Modelling
Advisor: Prof. Arie Kaufman

Virtual Shopping Experience project aimed to create a virtual environment in which a user can try out clothing before making purchases. My project on "3D Full Body Avatar Modelling with Microsoft Kinect" focused on creating a realistic 3D human model to be integrated in the shopping experience. I developed and proposed methods for point cloud modelling using Rigid and Non-rigid registration methods like Iterative Closet Point, Articulation and Features based registration. For gaming application we proposed to estimate human body parameters and fit the parameters in PCA space of a society of human models. (Implemented using Point Cloud Library, C++, Kinect SDK)


GPU Based Art Gallery and Watchman Route Problem for 2D and 3D Polygonal Domains: Variants and Applications
Advisor: Prof. Joseph S.B. Mitchell

The Art Gallery Problem (AGP) or the visibility coverage problem for guarding simple polygons or 3D polyhedron aims to find the minimum number of guards needed to guard the whole geometry. The AGP has been proven to be an NP-hard problem and hence most of the proposed algorithms aim to find a close to optimal solution in polynomial time. In this paper, we propose a novel and a sampling based algorithm for computing the guard number and their positions for a simple polygon with and without holes. The algorithm has been extensively tested on various classes of geometries and has shown very close to optimal results and remarkable performance. We also solve many variants of the problem and present an application for optimal light placement of 3D scene. The algorithm is further extended for finding optimal camera trajectory for target coverage for applications like creating virtual tours, generating street maps, etc.

(Paper Under Submission)


Facial Animation and Retargeting
Advisor: Prof. Hong Qin

Performance-driven animation aims to capture every trait and characteristic of an actor's facial and verbal expression and produce expressions on a digital double or creature. Creation of realistic digital faces is an important component in film and game production. In this project, I implemented a real-time and calibration-free facial performance capture framework using Microsoft Kinect sensor (RGB-D) data input. The expressions are re-targeted by estimating the weights of the linear blendshape model in PCA space.


Kitchen Scene Context based Gesture Recognition KSCGR
21st International Conference on Pattern Recognition (ICPR 2012), Japan

We propose a novel approach to a challenging problem of daily life cooking activity recognition task based upon object use and frame sequence tagging. We use a dynamic SVM-HMM hybrid model which combines structural as well as temporal video sequence information to jointly in fer the most likely cooking activity labels. We demonstrate that our approach can achieve activity recognition rates for kitchen scenarios of more than 72% on a real-world cooking dataset through kinect sensor (RGB-D format) consisting of 9 cooking activities with significant variations in performance of these activities by different subjects. Such a context based approach as discussed in this paper can be extended to other fine grain activities such as hospital operating rooms in medical practices, agricultural and manufacturing operations, etc.

Algorithm selected for on-site presentation for the contest on KGSCR as a part of ICPR-2012. I presented an online working demo using Microsoft Kinect at ICPR'12 Gesture Recognition Workshop organized by ChaLearn sponsored by Microsoft. Got 3rd prize.


Counter measures to 2D facial spoofing attacks
6th International Conference of Biometrics-2013, Spain

As a crucial security problem, anti-spoofing in biometrics, and particularly for the face modality, has achieved great progress in the recent years. Still, new threats arrive in the form of better, more realistic and more sophisticated spoofing attacks. The aim of this project is to recognize real face authentication and identify spoof attempts like print, video replays, etc. We used Linear Binary Pattern (LBP), GMM Fourier based Texture Analysis and Motion cues as potential schemes. The submitted results were evaluated on Replay-Attack database. Got 1st prize. Contest paper


Emotion Recognition in the Wild
Emotion Recognition In the Wild Challenge and Workshop EmotiW 2013
6th International Conference of Biometrics, 2013

Human Facial Expression is a very extensive problem in the field of computer vision and pattern recognition. In this work, we were given 380 video samples in the training set with their known expressions (including 7 categories: Anger/Disgust/Fear/Happy/Natural/Sad/Surprise) and in the test we had to classify 312 video samples under one of the above mentioned category of emotion. We used a blend of Optical flow, Gabor Filtering, facial keypoints, background motions and audio features for training the Support Vector Machine calssifier.



Enhancement of Foggy and Medical Images
Advisor: Prof. Anil K. Tiwari

Research work for information retrieval from degraded foggy and medical images using various schemes like Atmospheric Modeling for fog estimation, Retinex scheme, Local Contrast Enhancement, CLAHE, Wavelets, etc. Used Matlab.

SBUnix
Advisor: Prof. Mike Ferdman
Course: Operating Systems

Developed a Preemptive Operating System for Intel x86_64 architecture from scratch. Added features like COW Fork, Interrupt Handling, Virtual Memory Management, Ring Based Protection, Shell, Basic File System Support, System Calls etc. Code

Adding Extra Security Layer to GPU access:
Implemented the functionality to access GPU as a device driver in linux kernel with the help of UserModHelper API. This adds an extra layer of security as all GPU access is then passed through the kernel and need the root permissions in order to execute. Report



Point set registration via particle filtering and stochastic dynamics
Advisor: Prof. Allen Tannenbaum
Course: Medical Imaging

Implemented rigid and non-rigid registration of point clouds based on particle filtering in Matlab



Projects in Artificial Intelligence
Advisor: Prof. I.V. Ramakrishnan

Array of Artificial intelligence algorithms (informed state-space search, probabilistic inference, and reinforcement learning) implemented to automate Pacman game. Code



Click Modular Router & NDN Simulations on NS3
Course:Fundamental of Computer Networks

Click Modular Router: Used a network emulation testbed (emulab), and the click modular router for the performance evaluation of samplenetwork topologies under TCP and UDP.

NDN Simulations on NS3: Used and modified the ndnSIM module from the NS-3 network simulator to study the performance of a sample NDNarchitecture under different caching properties and traffic conditions.


Remote Sensing and GIS
Advisor: Prof. Sudhir Nagendra Gupta

Tracking LNMIIT routes and boundaries using GPS and carrying GIS overlay analysis to obtain mosaic image and then analyzing attributes (area/perimeter) of traced paths for (ILWIS 3.0) of LNMIIT using OpenJUMP and PostGIS


2013.
  • Shubham Bansal, Shubham Khandelwal, Shubham Gupta and Dushyant Goyal, "Kitchen Activity Recognition Based On Scene Context," Proc. of 20th IEEE Int. Conf. on Image Processing (ICIP) , 2013, Melbourne, Australia.

  • Dushyant Goyal et al., "The 2nd Competition on Counter Measures to 2D Face Spoofing Attacks," Proc. of 6th IAPR International Conference on Biometrics (ICB), 2013, Spain.

    2012.

  • S Mohanty, S Banerjee, J Jose, D Goyal, A K Mohanty and F Carminati, "Two dimensional correlated sampling using alias technique," Proc. of Journal of Physics, Computing in High Energy and Nuclear Physics (CHEP), 2012, New York, USA.

    2011.

  • Da-Yu Kao ,Shiuh-Jeng Wang, Dushyant Goyal, J. Liu, "A Trustworthy Computing of ADAPT Principle Guaranteeing Genuine Medical Image," Proc. of IEEE Int. Conf.on Parallel and Distributed Systems (ICPADS), 2011, Taiwan.

    2010.

  • Manoj Alwani, Dushyant Goyal and Anil K. Tiwari, "Medical Image Enhancement Algorithm Based on Local Contrast Enhancement and Human Visual Characteristics,," Proc. of the 7th Int. Conf. on Biomedical Engineering (BioMed), IASTED, 2010, Austria.

  • Manoj Alwani, Dushyant Goyal and Anil K. Tiwari, "A Wavelet Transform Based Medical Image Enhancement Using Human Visual Characteristics," Proc. of 2nd Int. Conf. on Digital Image Processing (ICDIP),, 2010, Singapore.
  • Dushyant Goyal

    Graduate Student

    Department of Computer Science
    SUNY Stony Brook University
    Stony Brook, New York - 11790

    Email: goyal1dushyant [at] gmail.com
    Phone: +1 631 551 3038

     
    Copyright © 2014 by Dushyant Goyal