Myriad is a fault-tolerant software library for the rapid development of ROI detection applications. The library bundles signal/image processing, machine learning, concurrency, and hardware acceleration all in a single, easy-to-use library for software developers targeting Java 8. A concurrent data-processing pipeline can be written in less than 250 lines of code.
In 2016 Emphysic was awarded a Phase I SBIR contract by NASA to develop the framework for the analysis of large Nondestructive Evaluation (NDE) datasets. The datasets are mapped to scores of parallel, independent processes running on a mobile device, standard computer, or a networked cluster of machines. Each process scans a subset of the data for indications of structural damage and as independent entities are unaffected by errors in fellow processes or system failures. If a process fails, only its work is lost as the system continues to process the data; the work lost is immediately picked up by another process. The results of the parallel analyses are compiled back to the original dataset with structural flaw indicators. Emphysic demonstrated training a machine learning model to detect structural damage in its Phase I Technical Review.
As part of the Phase I efforts for NASA two Graphical User Interface (GUI) applications based on the Myriad library were developed.
Trainer is a desktop application designed to ease the process of training machine learning models for Region Of Interest (ROI) detection applications.
Desktop is a desktop application that provides an easy to use graphical user interface for scanning datasets for Regions of Interest (ROI) built on Myriad.