About me
Hi 👋, I am currently pursuing my Master’s in the EECS Department at University of Michigan with my major in Image Processing and Machine Learning. I am broadly interested in the field of Deep Learning. Somethings I am currently excited about are devising scalable and generalizable algorithms which are able to work on real world data and the area of 3D Computer Vision.
Prior to this, I pursued my undergrad education in Electronics and Communication at National Institute of Technology, Karnataka, India. My Undergraduate thesis was in Copy Move Image Forgery Detection under the guidance of Dr. Sumam David. During my time as un undergrad, I also pursued an internship at Samsung Research Bangalore, where I worked towards making configurable SRAMs. I was also a part of IEEE - NITK Chapter and was involved in club activities such as going to a nearby college for a tutorial on Computer Vision and OpenCV.
Apart from this, I love reading novels fiction and non-fiction. A recent novel I read is Into Thin Air by Jon Krakauer. I also follow and play soccer, like astrophysics and talking about life in general. Feel free to reach out to me on my email.
Updates
[Jan 2021] Started as a GSI for Computer Vision by David Fouhey and Justin Johnson
[Sep 2020] Started as a GSI for Computer Vision by Andrew Owens
[Jun 2020] Started internship in the Perception & Prediction team at Uber ATG
[Sep 2019] Started as a Masters student at University of Michigan
[May 2018] Started as an Intern at Samsung Research, Bangalore
[Aug 2015] Started Bachelor’s at NITK
Projects
SICGAN
End to end Single Image Conditional GAN (SICGAN) framework for generating realistic meshes of 3D objects using a single RGB image. Were able to get more realistic looking meshses compared to Pixel2Mesh with a similar reconstruction error
Copy-Move Forgery Detection using SIFT and GLCM-based Texture Analysis
Algorithm based on SIFT and GLCM to detect multiple copy move forgeries in the same image where each image has undergone a different geometrical transformation and is robust to JPEG Compression.