Needless to say, research in these fields go hand-in-hand with a lot of experience in programming, algorithm design, and real-time systems, among other computer science essentials.
I currently hold a computer science Research Assistantship position at the GRASP lab @UPenn, under Dr. Ben Taskar.
[May 2012 - Present]
My current research with Dr. Ben Taskar and Ben Sapp is based on using machine learning and computer vision based techniques for Human Pose Estimation- in order detect important features of gait analysis. For this we are using a Kinect to record different gaits so that we can use its depth and RGB streams to perform pose estimation on the resulting point-cloud data. The pose estimation would basically consist of a learning algorithm that estimates in detail the human pose in a video using Pictorial Structures (i.e. by representing the human body as a structure made of a finite number of connected components.)
[Oct 2011 - April 2012]
Under Dr. Camillo Jose Taylor, this research work in the GRASP lab was concerned with the use of a network of low-cost, easily-deployable smart cameras in a scenario where sensors or devices need to be tracked efficiently.
The set-up consists of a network of multiple "smart" cameras that recognize certain pre-specified blink codes and communicate to each other wirelessly, thus forming a distributed tracking system, and a blinker circuit which I designed that would be recognized by these cameras.
The main objective of my research was to establish the correct parameters of a projective transform, or Homography, between the ground plane and the image plane of each camera, so as to facilitate effective sensor tracking when this blinker was placed on a mobile robot that would go through the viewing space of the cameras.
[May 2010 - July 2011]
This work involved the design of a real-time dynamic hand gesture recognition algorithm, implemented to recognize hand gestures using just the video input of a web-camera.
The algorithm uses Haar-like features and orientation histograms, and has been implemented using openCV and integrated with graphical applications like "slide-show", where the slides of a presentation could be moved by merely waving the hand, and "virtual drawing", in which one gesture (palm-up) is used to draw and another gesture (palm-sideways) is used to erase in an application similar to Paint. The work was demonstrated on an ARM-based Beagleboard, running Linux.
- Short-listed among top 20 projects at the Indian Institute of Science, Bangalore, India; in the national-level Jed-I Project Challenge(June 2011).
- The paper written on this work can be found here.
[October 2009 - June 2010]
Participated in the 18th Annual IGVC, held at Oakland University, Rochester, Michigan, from June 4-7, 2010. Involved building an unmanned ground vehicle capable of following a prescribed track while avoiding all obstacles in its path, and also having the functionalities of waypoint navigation (i.e. capable of autonomously using an in-built GPS to get to a prescribed destination co-ordinate). The vehicle was also capable of waypoint navigation using a programmed GARMIN HPS-17 GPS receiver.