Publications

  1. Kartikaeya Kumar, Lev Poretski, Jiannan Li, Anthony Tang. 2021. Tourgether360: Collaborative Exploration of 360° Tour Videos using Pseudo-Spatial Navigation (under review at ACM SIGCHI’22)

Past Research & Selected Projects

Visualizations for Survery Papers (Ongoing)

Supervised by: Prof. Haijun Xia, University of California San Diego • 2021

The goal of this project is to create visualizations that augment research survey papers. Usually, these papers cite hundreds of other papers that represent that particular subfield. Therefore, the information present in survey papers is quite dense to assimilate easily. Leveraging the sematic and syntactic structure present within the text, our goal is to create such visualizations that aid in the understanding of these research survey papers.

CrossDroneVR: A cross device controller for mobile VR

Personal Project • 2021

Traditional mobile-VR experiences like Google cardboard lack an input mechanism, rendering the user a passive observer. CrossDroneVR is a concept prototype in which one smartphone is used for displaying the experience in mobile Head Mounted Display, and another smartphone is used as a wireless control. CrossDroneVR is a physics-based First-Person View (FPV) drone simulation designed to simulate the thrill of FPV racing.

Tourgether360: Pseudo-spatial Navigation for 360 tour videos

Supervised by: Prof. Tony Tang, University of Toronto • 2021

Tourgether360 is a Unity Prototype that allows multiple people to watch a 360 video together. Users are represented as spherical avatars, which are embedded inside the video. These avatars move along a blue track in the environment depending on the user’s position in the video timeline. Users can point to specific locations using spatial markers, and in addition to the scrubber, users can navigate the video using an overhead minimap.

VRU-Net: A Lightweight CNN architecure for Human Pose Estimation

Supervised by: Prof. C. Krishna Mohan, IIT Hyderabad • 2020

Human Pose Estimation(HPE) is the problem of localization of human joints (elbows, wrists, etc) in images or videos. VRU-Net is an novel CNN architecture optimized for autonomous driving scenarios. Specifically, detecting and analyzing the pose and intent of Vulnerable Road Users (VRUs). The model has a VGG backbone and the main architecture is inspired from the Stacked Hourglass and the U-Net architecture.

AR-chitect: An Augmented Reality tool for architects

Supervised by: Prof. Pradeep Yammiyavar, IIT Guwahati • 2021

AR-chitect is a Unity prototype that projects CAD models in the real world in Augmented Reality, making it easier for architects to quickly compare designs. Architects can also create dynamic forms(like google forms) to elicit user ratings for different architectural designs in buildings.

Hand Tracking & Gesture Recognition

Electronics Club, IIT Guwahati• 2020

An Image processing-based hand gesture recognition system built through OpenCV that lets users play an FPS game through hand gestures.

Circuit-X: An Image Processing Challenge

Electronics Club, IIT Guwahati• 2019

Circuit-X is an algorithm based on Image Processing techniques that produce the truth table of a digital circuit given the Schematic diagram of the circuit.

Stopwatch/Timer in Verilog

Electronics Club, IITG • 2019

A structural code of a Timer/StopWatch in System Verilog HDL implemented on a BASYS-3 FPGA board.