Baskar Ganapathysubramaniam is Anderlik Professor of Engineering at Iowa State University. He is the director of the NSF/USDA AI Institute for Resilient Agriculture (aiira.iastate.edu). He directs a curiosity driven, computational sustainability group (me.iastate.edu/bglab) with research interests in the areas of scientific computing, applied mathematics, and machine learning with applications in food, energy, and healthcare systems.
Extracting temporal changes in plant phenotypes during the growing season (for instance, changes with time of height, biomass, and number of leaves) can provide valuable information about crop genetics, development and physiology. However, multi-time point, high throughput phenotyping under field conditions is challenging, even with state-of-art machine learning (ML) approaches. A key bottleneck is the difficulty of creating large scale annotated data that is a prerequisite to training ML models. This talk will illustrate domain adaptation and sim2real approaches that alleviate the need for large scale annotation, and leverage indoor or available datasets. This is collaborative work with several faculty in the AIIRA team.