*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Title: Framework for Cloud Formation Classification in a Continuous Active Learning Paradigm
Committee:
Dr. Wills, Advisor
Dr. Blough, Chair
Dr. Hao
Dr. Kira
Abstract: The objective of the proposed research is to investigate an open and interesting research area in computer vision and machine intelligence – cloud formation classification. This is closely tied to increasing attention to climate change and its effects – an alarming decline in natural resources, dilapidation of natural habitats and its inhabitants, and highly irregular, more frequent climate-related disasters – threatening global health and our immediate future. There is increasing demand for robust cloud formation data to help mitigate expensive climate modeling and to provide sufficient information in environment-aware applications, such as path planning in UAVs on critical missions. Although a few recent ML-based vision approaches for interpreting cloud formation data have performed to an acceptable level, there is a lack of sufficient and robust datasets on cloud formation because of varied labeling vocabulary and limited expert labeling resources. This lessens the quality of information available for effective research work in this area. The proposed work focuses on establishing a framework for cloud formation classification in a continuous and active learning paradigm, which builds a robust labelled set of trusted cloud formation data, curated from data collected from various untrusted sources, including citizen-scientists and enthusiasts. This framework will also utilize identified features represented in collected data to provide near-real-time classification reasoning to contribute to the training of citizen scientists and amateur enthusiasts to better categorize cloud formation with little to no expert supervision. The application of this framework will enable a robust method for collecting cloud formation data for cloud modeling and climatic change research, aviation path-planning, and other efforts that contribute towards educating society to become more environmentally aware.
