Self-Supervised and Continual Representation Learning in Unstructured Environments

Speaker:  Peyman Moghadam – Brisbane, QLD, Australia
Topic(s):  Artificial Intelligence, Machine Learning, Computer Vision, Natural language processing

Abstract

In today's robotic applications, learning-based methods play a key role in tasks like (re)-localization, perception, navigation, and manipulation. However, these methods often face challenges when generalizing to new environments due to domain gaps and the scarcity of large-scale, labelled data. This lecture will explore alternative paradigms in which robots incrementally learn representations from sequences of environments while addressing the issue of catastrophic forgetting. It will delve into strategies for adapting pre-trained large models in the face of significant domain shifts, using self-supervised methods that do not rely on ground-truth supervision.

About this Lecture

Number of Slides:  30 - 40
Duration:  45 - 60 minutes
Languages Available:  English
Last Updated: 

Request this Lecture

To request this particular lecture, please complete this online form.

Request a Tour

To request a tour with this speaker, please complete this online form.

All requests will be sent to ACM headquarters for review.