AG2PI Workshop #5 - Phenomic Data Processing Using Machine Learning and Distributed Computing смотреть онлайн

AG2PI Workshop #5 - Phenomic Data Processing Using Machine Learning, Distributed Computing & Container Technology
July 22, 2021

This hands-on workshop will show how to leverage our pipelines to process your proximal and remote-sensing data. We will showcase two pipelines that leverage distributed computing, machine learning, and container technology. First, we will run our thermal pipeline, which detects individual plants in thermal images and runs Kmeans clustering to collect individual plant temperature values. Then, we will run our hyperspectral pipeline, which detects Spectralon reflectance reference targets and calculates plot-level reflectance values.

The University of Arizona is home to the world's largest phenotyping platform, capable of scanning a quarter-mile long agricultural field using a combination of sensors including: a laser line scanner; hyperspectral imager; and thermal infrared, RGB, and fluorescence cameras to generate large, multimodal datasets. Collectively, these sensors capture up to 10 TB of data per day, which overwhelms the computing capacity of most research institutions resulting in a significant data processing bottleneck. PhytoOracle was developed to address two key challenges of phenomic analysis pipelines: (1) accelerating analysis tasks by integrating distributed computing resources and managing high throughput data, (2) containerization of computational code for improved ease-of-use and reproducibility.