"Neural Network Processing of Holographic Images"
SPEAKER: Dr. John Schreck - NCAR/CISL
DATE: April 19, 2022
Time: 3:30 - 4:30 pm MST
WEBCAST: https://operations.ucar.edu/live-eol
QUESTIONS: Slido will be used to answer your questions during the seminar.
______________________________________________________________________________________________________
ABSTRACT
John will present results from a promising new machine-learning based paradigm, HolodecML, for processing field-acquired cloud droplet holograms. Holographic images are used as input to a neural network model that predicts masks around in-focus particles, from which the position, diameter, and shape are computed. The approach is fast, scalable, and takes advantage of GPUs and other heterogeneous computing platforms. It combines applications of transfer and active learning by using synthetic data for training, and a small set of hand-labeled data for refinement and validation. Artificial noise and transformations applied during synthetic training enables optimized models for real-world situations.
For more information or questions, reach out to Jacquie Witte.
