HOLODEC ML

Term: 2020 - 2025

Funding source: NSF

Partners: NSF NCAR Earth Observing Lab

HOLODEC is an airborne cloud particle imager that collects high resolution images multiple times per second on flights through clouds during field campaigns. By virtue of imaging the particles with a laser before they make contact with the aircraft, the HOLODEC probe can capture the 3D distribution of cloud droplets and ice crystals with amazing precision. The challenge with HOLODEC is that inverting the 2D images back into 3D particle locations and sizes is very computationally demanding.

The MILES group has developed deep learning algorithms to process HOLODEC holograms very quickly on GPU and can accurately locate and characterize individual particles. We are currently in the process of analyzing a whole field campaign’s worth of HOLODEC images.