Measuring coastal sediment grain size instantly using Instagrain, a hand-held camera with on-device machine learning

Much of the US coastal zone is covered in a veneer of mobile sediment, and the size of this sediment determines under what conditions it moves. Accurate measurement of coastal sediment grain size is critical for work on coastal erosion, shoreline change, total water level forecasting, storm impact predictions, planning, designing nature-based coastal protections, and other coastal resilience projects. Measurement of grain size is time consuming, costly, and requires a laboratory. Inaccurate measurements result in degraded performance of models, plans, predictions, and forecasts. To address slow measurement speed and high cost-per-sample, we have built a new handheld camera-based system that uses on-device machine learning, does not require calibration, and provides accurate field measurements of grain size within 1 second, ~1,000,000x faster than lab quotes of 2 weeks. We propose three research objectives to dramatically improve the performance of our technology for coastal sites of the US, regardless of sediment size, distribution, color, and composition: First, field collections to grow our training dataset; Second, machine learning model development to increase performance and reduce training data requirements; Third, targeted work to determine presence and percentage of specific minerals used for some coastal projects.