AI-enabled processing of sensing data to extract trait values for use by breeders
Year 5 Research Project

AI-enabled processing of sensing data to extract trait values for use by breeders

Description

Affordable sensors on ground rovers and drones can measure crop traits at scale, but processing the data and extracting traits remain a major bottleneck for plant breeders. This project continues the development and deployment of GEMINI (Genotype x Environment x Management Innovation in Intelligence for climate adaptation), an open-source, browser-based app that lets breeders process field sensing data without writing code. Users can adjust plot boundaries, also in fields without GPS-enabled planting, and run automated trait extraction such as flower and pod detection. The app already includes several years of field data from multiple crops and locations. The project also uses the Helios 3D biophysical crop modeling framework to generate synthetic thermal images, and trains machine learning models to estimate stomatal conductance parameters from them. This links sensing data to crop models, breeding and genomics. We will publish the app, offer live and online training, reduce the need for manual annotation through synthetic data, and report model performance for tested and untested genotypes and environments.

Team

Portrait of Mason Earles

Mason Earles

Principal Investigator

Portrait of Brian Bailey

Brian Bailey

Co Principal Investigator

Portrait of Christine Diepenbrock

Christine Diepenbrock

Co Principal Investigator