We use data, AI, and computing to understand our planet and how we can live better on it.
Lab News
May 2024: Our paper benchmarking vision-language models (VLMs) on Earth observation data won the Best Paper Award at the ICLR ML4RS Workshop. Congratulations Chenhui! Benchmarking reveals that VLMs like GPT-4V are good at scene-level tasks (location recognition, land cover classification) but bad at counting and localization.
February 2024: Our work mapping crops in Thailand, led by PhD student Jordi Laguarta, was featured on MIT News and MIT’s front page! This is the first time a crop type map has been created for Thailand at a field level, and our work provides a template for mapping crops in smallholder regions where data has historically been unavailable.
January 2024: New work out in Science! The Clean Water Act is the main US environmental law that protects waters from pollution and development, but which waters are protected have never been mapped. In a collaboration with UC Berkeley, we used deep learning and geospatial data to understand which water bodies are regulated by CWA.