The Center for Geospatial Information Science (CGIS) presents a talk by Assistant Professor Gengchen Mai from the University of Texas at Austin on "Geo-Foundation Models and Spatial Representation Learning."

The event will take place from 11 a.m. to 12:30 p.m. on Friday, Oct. 4 in Room 325 in River Road. Zoom will be available for remote attendees.

Mai will discuss the application of foundation models—large-scale neural networks adapted for various tasks—to geospatial research. His recent work includes using these models for tasks such as extracting location data from social media posts, predicting sustainability indexes, and image geolocalization. Mai will also address the challenge of current models’ inability to handle geographic coordinates, proposing spatial representation learning (SRL) as a solution. SRL aims to enable neural networks to learn from spatial data, such as points and polygons, in their native formats. He will also introduce TorchSpatial, a framework for location encoding.