Creative Scientific Collaborations using Deep Learning emerge from COVID Disruptions

We are very proud to have been able with work with colleagues are Argonne National Laboratory to team up with some of their staff, whose usual activities were disrupted by COVID-19, in data labeling for deep learning modeling.  The team was led by Ryan Jacobs (shown above) at UW and included:

UW-Madison: Ryan Jacobs, Dane Morgan

ANL: Ben Blaiszik, India Gordon, Monica White

Here is a brief description:

Creative solutions to combat disruptions from COVID-19 have emerged through an innovative scientific collaboration between administrative staff at Argonne National Lab (ANL) and researchers at the University of Wisconsin-Madison. The collaboration used open-source web application software to create a large dataset of labeled microstructural images of steel alloys used in nuclear reactors. This dataset was then used to train  state-of-the-art deep learning object detection methods to automatically label key image features critical to understanding the material performance in a reactor environment. This synergistic combination of open source software, data infrastructure, and deep learning methods applied to understanding steel alloy behavior for nuclear reactors will directly impact materials property modeling and design considerations for materials used in nuclear reactors, a key pillar of US energy infrastructure.