Case Study: Harte Research Institute improves shoreline classification accuracy with Google for Education AutoML Vision

Google's machine-learning tools help researchers assess and track environmental change

Google for Education helped the Harte Research Institute for Gulf of Mexico Studies at Texas A&M University - Corpus Christi tackle a complex environmental challenge: classifying aerial shoreline imagery to track coastal change and support conservation and oil-spill response planning. The team needed a more efficient way to assign Environmental Sensitivity Index (ESI) values across nearly 9,000 miles of Texas coastline, a process that had traditionally required expert manual review.

Using Google Cloud’s AutoML Vision, Google for Education enabled the Harte Research Institute to build custom image classifiers from its own shoreline data, including multi-label datasets for more accurate results. The solution improved model performance from 84% precision and 78% recall with single-label training to 91% precision and 90% recall with multi-label training, making it much easier for non-experts to assign ESI values and scale predictions through an API.

Harte Research Institute

Anthony Reisinger

Assistant Research Scientist