Initial Demonstration of Self Supervised Learning Project

Drew Bollinger from Interagency Implementation and Advanced Concepts Team (IMPACT) demonstrated a prototype of the self-supervised learning (SSL) based image similarity search tool to the Earth Science Data and Information System (ESDIS).  The SSL project was sponsored by IMPACT as a 2020 Frontier Development Lab (FDL) Earth science challenge to address a need within Earth science data archives to search for similar images using a user provided query image.  The FDL team continued to develop and enhance the SSL model post FDL challenge period with the goal of infusing the model into ESDIS enterprise services such as Worldview visualization tool.  To that end, Mr. Bollinger demonstrated scaling the model to work with large scale image archives and conducted feasibility studies using commercial cloud computing infrastructure.  ESDIS and IMPACT plan to infuse this technology with respective tools.

IMPACT logo
Scroll to Top