Research Committee, Working Group 2

Initiative on Developing Open and Reusable GeoAI Tools

Members

  • Yingjie Hu (Leader), University at Buffalo
  • Jochen Albrecht, Hunter College
  • Wenwen Li, Arizona State University
  • Wei Luo, Harvard University
  • Ming-Hsiang Tsou, San Diego State University
  • Lynn Usery, U.S. Geological Survey
  • Eun-Hye Yoo, University at Buffalo
  • Lei Zou, Texas A&M University

Geospatial Artificial Intelligence (GeoAI) provides novel approaches for addressing a variety of problems in the natural environment and our human society. While a lot of research progress has been made, there is currently a lack of open and reusable GeoAI tools shared on publicly accessible repositories. Consequently, it is difficult for GIS researchers and educators to find and reuse these developed tools. In addition, since the performance of deep learning models is often affected by implementation details, such as random seeding and parameter initialization strategies, it can be challenging for researchers to re-implement and reproduce the same tools by only following the methodological descriptions in their papers. In this research initiative, we propose to encourage the development and sharing of open and reusable GeoAI tools and organize them into a collection. This initiative can contribute to both GIScience research and education. For research, it can become easier for researchers to find and reuse existing tools, or to compare a newly developed method with the existing ones. For education, this initiative can help instructors teach GIS courses related to GeoAI by demonstrating existing GeoAI tools in a way similar to how we teach some more traditional GIS operations, such as buffer and clip.

The objectives for the Developing Open and Reusable GeoAI Tools initiative are as follows:

1. To develop an initial collection of GeoAI methods with implemented software tools shared on publicly accessible repositories. We plan to organize a workshop in conjunction with the 2020 UCGIS Symposium. The workshop submissions will be peer-reviewed and will be published in a book or a journal special issue. Each published paper will be accompanied with an implemented GeoAI tool that is publicly accessible.

2. To promote GeoAI research and education. The developed GeoAI tools can be re-used by other researchers to build new methods, and a GIS educator should be able to follow the descriptions of a developed tool for teaching and demonstration purposes. 

3. To encourage further developments of GeoAI methods and tools. This initiative will nevertheless generate only a small collection of tools. However, we hope this initiative can encourage the development of more GeoAI methods and tools in the following years.