Part 1 will focus on the use of Doodler (https://github.com/Doodleverse/dash_doodler
), a 'human-in-the-loop' labeling tool for image segmentation (described in this paper: https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2021EA002085
). We'll cover the two primary uses of Doodler; a) for relatively rapid image segmentation of a small set of images, and b) for making libraries of labeled imagery for training Machine Learning models to automate the process of image segmentation on larger datasets. We'd ideally like participants to label the same imagery in-class so we can discuss image interpretation and label agreement. This may even result in a publishable dataset; participants would receive co-authorship and could opt-in/out.
We will provide example datasets and models, but participants will also be encouraged to bring their own imagery sets. That way, participants will have time to familiarize themselves with the burgeoning Doodleverse tools (https://github.com/Doodleverse
) in between classes on their own data.