About
Hi! I'm Sreeharsha Paruchuri, an Senior Year Undergraduate Researcher at the Robotics Research Center, International Institute of Information Technology - Hyderabad, where I'm pursuing a degree in Electronics and Communication Engineering with honors in Robotic Vision.
As an avid coder, I thoroughly enjoy learning new things which help me acquire a broad skillset, enabling me to tackle any problem statement. During my years at college, I learned about the prowess of what code could do which motivated me to take on projects in multiple domains which I believe has helped me become a well-rounded software engineer . I hope to be able to employ this skill in creating software which will impact the lives of others be it from autonomizing to improving the efficiency of tasks that we take on as human beings.
I'm also piqued by the limitless applications of Artificial Intelligence, more specifically Deep Learning and Computer Vision. The intersection of the latter with Robotics is an area of research that I enthusiastically work on as advised by Dr. Madhava Krishna..
"Youtube is the closest thing we'll ever get to a time machine."
, The influence of music in our daily lives can not be embellished enough, the quote builds on the human tendency to assosciate music with tendencies, experiences and people. Working alongside Dr. Vinoo Alluri, my research explores the facets of music and music listening that are characteristic of individuals suffering from mental maladies. Modern Machine Learning and Statistical techniques enable us to perform rigorous testing to validate our hypotheses and uncover latent features of music listening that can be used as predictors for the risk of depression for an individual.
Sports, Photography, Writing, Music and Reading are just a few of the hobbies I enjoy. I primarily play Cricket, Table Tennis and Volleyball. I believe that photography can tell stories that words can't and relish capturing nature's fleeting moments, among others.
I am also a fanatical car enthusiast, so it might seem somewhat hypocritical of me to say that I am also an avid environmentalist. I actively follow advances in technology (ranging from mobile phones to spacecraft). On a slightly tangential note, I find the uncertainity of the questions that come with philosophy to be very contemplative, more so the age old mind-body debacle and investigating nature through a teleological lens. I'm not incredulous to hearing out what others think about the same so feel free to hit me up if you'd like to give your two cents on any of the above 'Meditations' :)
I look up to Steve Jobs, Elon Musk for their remarkable journeys and what they stand for.
Major Projects
The only problem without a solution is an improperly phrased one.
Do check out my GitHub for more projects / codes.
Dense Stero Reconstruction in real-time
Given a set of stereo images from the KITTI dataset, we undistort and rectify the stereo pair. We employ epipolar geometry to find correspondences between the images and then use block-matching to compute a disparity map for each stereo pair. Poses are estimated using ORB-SLAM 3 and the sequence is registered and visualized as a point cloud and an octomap.
Neural Machine Translation
Comparing an LSTM and GRU based Bahdanau Attention architecture (as described in this paper ) for a popular NLP Sequence2Sequence task of Machine Translation. Both approaches make use of the ADAM optimizer. The resulting English to Hindi translation is evaluated using the BLEU score.
Pose-Graph Optimization
This project deals with a popular backend optimization framework for SLAM techniques . This SLAM backend uses information from the frontend to model the evolution of the robot in the enviornment as a graph. A non-linear least squares optimization is done using the jacobian of residuals which includes weighted loop-closures. Iteratively, an optimal pose-graph is solved for. A paper on the Past, Present and Future of SLAM is discussed.
Artpark Robotics Challenge
Qualified to the second stage of this international competition which called forth a seamless integration of computer vision, navigation, dynamics and control algorithms in a modular software stack. Contributed to the computer vision algorithm to detect trash items and a trash can in real time using data from a RGB-D stereo camera. Used the navigation stack in ROS to communicate waypoint messages to the multiple nodes and SLAM backend.
Teaching
If you want to master something, teach it
These are words from the legendary Richard Feynman, a person who profoundly motivates me to find joy in learning something new with every experience. I have have been fortunate enough to have the opportunity to be a proponent of knowledge by working alongside Professors as a Teaching Assistant for multiple courses at IIIT, Hyderabad. Not only does teaching strengthen your conceptual understanding of a topic, but it is a form of service to others, a chance to be helpful by giving back. I was able to contribute to the following courses by teaching, guiding and evaluating students by taking tutorials and setting assignments.
Frames
To be honest, I was at the right place at the right time
Blogs
Truth be told, I've never fancied writing. But, I've always enjoyed reading, be it fiction or non-fiction. If you allow yourself to immerse yourself in a book for those few hours, you are not you. You see the world from someone else's shoes, experiencing events in all their glory and tragedy while possibly learning from them.
It was when I was chosen to be a panelist for a moderated talk that I discovered a passion for writing, be it narrating a story or offering a well-researched and erudite opinion on a topic. At Ping! (the independent student media body of IIIT-H) I serve as a writer and an editor.
A couple of my most recent articles:
The Psychology Behind Why Humans Find it Hard to Stay Home : A short feature on how humans have come to be the social creatures that we are today and relate that to how lockdown has left most of us in a state of lassitude.
An Online Felicity: Exploring challenges and possibilities : With the outbreak of COVID-19 immuring students to their houses, we investigate possibilities as to how Felicity (the annual cultural fest of IIIT-H) could shift online.