GSA Connects 2021 in Portland, Oregon

Paper No. 245-8
Presentation Time: 3:20 PM


SUN, Ziheng, Department of Geography and Geoinformation Science, George Mason University, 4087 University Dr Ste 3100, Fairfax, VA 22030-3415, ALNAIM, Ahmed, Center for Spatial Information Science and Systems, George Mason University, 4087 University Dr, 3100, Fairfax, VA 22030 and MA, Xiaogang, University of Idaho, 875 Perimeter Drive, MS 1010, Moscow, ID 83844-0001

AI/ML is having a long prospective development in almost every corner of the science community. However, making AI/ML actually work in solving geoscientific problems is still problematic and troublesome. One of the major obstacles is the high barrier of entry and the huge burden laid on the shoulder of AI practitioners. Although it is relatively easy to learn the off-the-shelf AI/ML models, it is a big obligation to operationally construct and manage a working full-stack AI workflow, containing all the steps from pre-processing Earth observations into AI-ready format, train and test and deploy the models, and post-processing the model results into useful information. To make full-stack AI workflow easy to manage and maintain and lower the barrier of entry for geoscientists to get on board, we present a software, GeoWeaver, which can help people to intuitively compose full-stack AI workflow, execute and manage the provenance, accelerate the transition from research to operation, minimize the coding efforts, and eventually open a new door to AI for geoscientists, who think they are NOT fit for AI, to utilize AI techniques in their research easily and successfully.