Open Source Parallel Image Analysis and Machine Learning Pipeline

Today, NASA researchers must create, debug, and tune custom workflows for each analysis. Creation and modification of custom workflows is fragile, non-portable and consumes time that could be better spent on advancing scientific discovery. The Phase I open source software Ensemble Learning Models (ELM) provides composable, portable, reproducible, and extensible machine learning pipelines with easy-to-configure parallelization, with tools specifically for satellite data processing, weather and climate data processing, and machine learning and prediction. This is a major advancement over the current state-of-the-art because of reduced workflow creation time, parallelization, portability of deployment and use, extensibility, and robustness. Phase II will extend the Phase I work with more options useful to NASA missions, such as advanced ensemble fitting and prediction tools, feature engineering options for 3-D and 4-D arrays, and a web-based map user interface. Phase II will also harden and extend ELM to make ELM's easy-to-use large data ensemble methods accessible to industry outside of NASA, increasing the potential user base in a variety of domains.

Continuum Analytics sees direct usage applications in any NASA project deriving analytical value from multidimensional climate data arrays and hyper- or multispectral remote sensing data, such as climate reanalysis, landscape change analysis, land cover mapping, or drought or vegetation indices. In Phase I the team provided a number of scientific data loading tools for formats common in NASA remote sensing and climate science missions. Phase I work also created parallel ensemble fitting and prediction methods for a variety of unsupervised and supervised machine learning models. Phase II will provide tools for NASA data formats, additional tools for feature engineering in multidimensional climate data arrays, and advanced options for ensemble fitting and prediction, like hierarchical modeling and vote count ensemble averaging. Phase II will also include work on a web-based map interface and demonstrations of how ELM MLT may be useful in NASA missions like climate reanalysis and land cover classification.

The team sees direct usage application of the image analysis and machine learning pipeline outside of NASA, such as: - NOAA mission-related research to predict changes in climate, weather, oceans and coast, and conserving and managing coasting and marine ecosystems and resources. - DOD/IC - foreign defense and homeland security applications - Commercial infrastructure and engineering, disaster management and mitigation analysis, natural resource monitoring, energy-related exploration and operational management. - Flood and floodplain mapping for insurance adjustments, bridge construction projects, FEMA floodplain definitions, river habitat and restoration projects, and emergency planning at local, state, and federal agencies - Forest disease and insect damage density identification for large commercial forest owners - Snow and ice cover and recession analysis useful in climate change and water management planning at federal, state, and local agencies - Developing spectral identifiers of agricultural crops in healthy versus water and nutrient stressed conditions - Classifying parking lots and roads based on the number of vehicles evidently in the image, an indicator of economic activity and also potentially useful in federal security applications - Mapping ecologically sensitive and geotechnically unstable areas, such as wetlands and mass wasting events, useful for reducing the cost of development review in local, state, and federal environmental agencies, remote asset trackin

This is early/mid-stage (TRL 6) — the most realistic path in is NASA SBIR/STTR, which funds small businesses and research institutions to develop technology aligned with NASA's needs (equity-free, phased funding). Check whether a current SBIR/STTR solicitation topic overlaps with this project's technology area, or contact the project directly (above) to ask.

None of these are guaranteed paths for this specific project — TechPort itself doesn't have an "apply" button. Reaching out to the contact(s) above with a specific question is usually the fastest way to find out what's actually open.