"Imitating Shortest Paths in Simulation Enables Effective Navigation and Manipulation in the Real World"
Thursday, Jan. 18th @ 11am
FAH 3002 and Zoom: https://ucsd.zoom.us/j/95057481896
Speaker: Tanmay Gupta
Can we train real-world robotic policies without RL, human supervision, or even real-world training? Could exploration emerge from imitating shortest-path planners? Can powerful visual encoders help bridge the sim2real gap? How efficient is the training process? How does the performance scale with the number and diversity of training episodes? What are the crucial architecture design choices? In this talk, we will answer these questions and more!
Bio:
Tanmay Gupta is a research scientist in the PRIOR team at the Allen Institute for Artificial Intelligence (AI2). Tanmay received his PhD from UIUC, where he was advised by Prof. Derek Hoiem and closely collaborated with Prof. Alex Schwing. Tanmay has received the CVPR 2023 Best Paper Award for his work on Visual Programming; Outstanding Reviewer Awards at ICCV 2023, CVPR 2021, and ECCV 2020; and served as an Area Chair for CVPR 2024, CVPR 2023, and NeurIPS 2023.