Imaging Science Ph.D. Defense: Manoj Acharya
Ph.D. Dissertation Defense
Towards Multimodal Open-world Learning in Deep Neural Networks
Manoj Acharya
Imaging Science Ph.D. Candidate
Chester F. Carlson Center for Imaging Science, RIT
Advisor: Dr. Christopher Kanan
Abstract:
Over the past decade, deep neural networks have enormously advanced machine perception, especially object classification, object detection, and multimodal scene understanding. However, a major limitation of these systems is that they assume a closed-world setting, i.e., the train and the test distribution match exactly. This means, for example, that an input that is from a category the system has never seen during training will not be recognized as unknown. However, many real-world applications often need this capability, e.g., self-driving cars operate in a dynamic world where the data can change quickly over time due to changes in season, geographic location, sensor types, etc. We must make systems that operate in an open-world, where the test distribution does not match the train distribution. In open-world learning, the system needs to detect novel examples which have not been seen during training and continually update the system with new knowledge, without retraining from scratch. In this dissertation, I address gaps in the open-world learning literature and develop methods that enable efficient multimodal open-world learning in deep neural networks
Intended Audience:
Undergraduates, graduates, and experts. Those with interest in the topic.
To request an interpreter, please visit https://myaccess.rit.edu
Event Snapshot
When and Where
Who
This is an RIT Only Event
Interpreter Requested?
No