Research
-
Selecting Source Tasks for Transfer Learning of Human Preferences
h. nemlekar · n. sivagnanadasan · s. banga · s. gupta · s. nikolaidis — icaros lab, usc
Preference-transfer learning in human-robot interaction: selecting which source tasks best transfer a human's learned preferences to tasks a robot hasn't seen.
-
Sketch2Code: Transformation of Sketches to UI in Real-time Using Deep Neural Network
v. jain · p. agrawal · s. banga · r. kapoor · s. gulyani — arXiv:1910.08930 (2019) · widely cited
A deep neural network trained on a custom sketch dataset that detects UI elements in a hand-drawn wireframe and generates UI skeleton code in real time across platforms.
// award — a project with a prize attached
first prize — marconi society celestini program india · 2019
Rakshak
An Android app for women's safety that detects distress speech patterns via the smartphone microphone — help/"stop" commands in distressed tones — then triggers an SOS with the user's location to pre-set emergency contacts. Built from the Google Speech Command dataset, extended with safety-specific commands. Published on Google Play.
team: piyush agrawal · subham banga · aniket sharma · ujjwal upadhyay — bvcoe, new delhi
program partners: iit delhi · anchored by dr. aakanksha chowdhery (then google brain) · marconi society, chaired by vint cerf
// education
-
MS, Computer Science — University of Southern California
robotics research, icaros lab · advisor: prof. stefanos nikolaidis · listed publicly as ms 2023
-
BTech, Information Technology — Bharati Vidyapeeth's College of Engineering
new delhi