News Article

NOAA Awards Nearly $700,000 to Enterpreneurial Machine Learning Projects
Date: Aug 26, 2020
Author: Oliver Peckham
Source: EnterpriseAI.news ( click here to go to the source)

Featured firm in this article: Kraenion Labs LLC of Los Gatos, CA



In the computing sphere, the United States' National Oceanic and Atmospheric Administration (NOAA) may be most well-known for its massive weather and climate models, which predominantly run on correspondingly massive supercomputers and clusters. With the advent of machine learning and artificial intelligence, however, lighter-weight applications are offering serious deliverables -- and receiving considerable funding. Now, NOAA has announced that it is awarding grants to 21 small businesses through its latest round of Small Business Innovation Research (SBIR) program funding, including five businesses working to improve NOAA's operations using machine learning.

The SBIR program targets the entrepreneurial sector, with NOAA explaining that "the risk and expense of conducting serious R&D efforts are often beyond the means of many small businesses" and SBIR loans -- capped at $150,000 per awardee -- can help those businesses to compete while promoting innovative research.

"As NOAA continues to strengthen its commitment to protecting life and property, we are increasingly reliant on the expertise and agility of the private sector," said Neil Jacobs, Ph.D., acting NOAA administrator. "Through collaboration with these small businesses, Americans will benefit with increased forecast accuracy and better management of our natural resources."

For the first phase of its 2020 awards, the program received 76 applicants, 21 of which will be the recipients of a total of $3.1 million in awards. Nearly $700,000 of that is going to five machine learning-enabled projects.

Philadelphia-based software and analytics firm Azavea received a full $150,000 award for its project, "Advancing Flood Extent Delineation Modeling Using Synthetic Aperture Radar (SAR) Data." Using the grant, Azavea will work to resolve a key problem with timely responses to flood events: seeing through the heavy cloud cover that often accompanies flooding. To combat the cloud cover, Azavea will apply synthetic-aperture radar (or SAR), which uses radar to reconstruct images and landscapes, in combination with deep learning techniques to interpret SAR imagery in real-time. "By combining these two technologies," Azavea writes, "this project will support the rapid delivery of accurate flood inundation maps that will enable first responders, humanitarian relief organizations, and other decision-makers on the ground to effectively route resources and identify highly impacted areas, both during and following extreme weather events."

The second-largest award of the trio -- just a few dollars short of $150,000 -- went to Kraenion Labs. The Los Gatos, California-based AI startup describes itself as "developing machine learning and active learning technology to analyze large 2D and 3D multi-spectral datasets of importance to public safety and national security" -- and using the grant, it'll be putting those talents to use in risk assessment. Using global satellite data in conjunction with Kraenion's proprietary "Vision Engine" platform, which prioritizes high-value samples for human-assisted labeling during the training process, the company proposes developing two tools. First, a property risk assessment tool aimed at the home insurance industry; second, a tool for coastal change analysis, land cover mapping and annotation of weather data in global models.

The Ann Arbor-based Michigan Aerospace Corporation squeaked in just twenty shy of that with a $149,978 grant. Michigan Aerospace's grant will be used to improve data accessibility. "NOAA's mission is critical," the company wrote. "We must find better ways to make NOAA's data more accessible and usable to educators who are teaching our next generation of decision-makers. For the value of the data to be recognized, it must be used -- and not just by data scientists." To that end, they propose the OPEN WORLDS NOAA Portal, which would build on the company's PLAIT.AI platform (used to monitor watershed health), allowing users to "ingest, process, visualize, and apply machine learning tools to NOAA data."

SafetySpect, a food contaminant detection company based in Los Angeles, received over $149,200 for rapid detection of fish species and quality in marketplaces. The company sees a pathway for machine learning to assist in combating fraud and deception in seafood marketing, wherein merchants sell one fish but claim it is another (typically more expensive) species. "We will demonstrate scalability and evaluate accuracy of machine learning classification algorithms for the assessment of fish quality, nutrient content, and species authentication," the company wrote. "This technology has the potential to be used as a rapid species identification method in restaurants, seafood markets, and other points along the seafood supply chain."

Finally, INNOVIM, a Maryland-based IT firm, was awarded $91,520 to develop a new machine learning technique for forecasting extreme precipitation from landfalling atmospheric rivers. These atmospheric rivers -- bands of concentrated moisture high in the air -- are responsible for large quantities of rain, offering both boons and dangers to the environments they deluge. However, because the rivers are concentrated, it is relatively difficult to forecast them more than a few days out. INNOVIM aims to use machine learning to provide an alternative avenue for forecasting this heavy precipitation, bypassing heavyweight weather models and instead analyzing the core datasets with an observational approach.

"We are excited about this year's awardees, many who are harnessing the power of unmanned systems, artificial intelligence, genomics, machine learning and public engagement to develop products and services that support NOAA's mission and may also have great potential as commercial products," said Kelly Wright, director of the NOAA Technology Partnerships Office.

This story first appeared on sister website Datanami.