Graduate Research Assistant
Published in Arizona State University, 2023
- Collaborated with MLCommons and NASA Harvest to design globally diverse benchmarking datasets for crop field boundaries using DataPerf standards.
- Designed a pipeline for creating and matching stratified samples of crop parcels to standardized EO images, including download, conversion, and patchlet sampling.
- Optimized cross-validation performance on boundaries with K-fold splits based on morphological characteristics.
- Evaluated baseline models MaskRCNN, Segment Anything Model (SAM), and UNET to improve a benchmarking accuracy for sparse crop field annotations.