AI-Enabled Change Detection Non-Traditional Events
What is the objective?
Desire to source and build relevant data sets in order to deploy artificial intelligence (AI-Enabled) software that can detect changes in objects, over time, in overhead imagery/video. This software should be flexible enough to detect changes in buildings/terrains/geographies/vegetation levels, in various lighting and weather conditions.
What problem are we trying to solve?
As the world economies continue to develop and natural resource usage rates change, natural and built environmental dynamics will impact the successes or failures of various world organizations. Even in current events, we can see how resource allotment, draught, flooding, and related environmental changes are affecting organizations, governments,countries and their populations. History shows that when a country’s resource needs are negatively impacted by natural or human-driven environmental changes, wars start.
This effort is intended to anticipate how non-traditional events, (in the eyes of computer vision), such as those stated above affect organizations and enable decision makers to react accordingly. The goal is to bring machine speed to the decision-making process. To make this happen, we need to:
• Ensure our data sets are complete.
• Enable acceptance of new data to this set.
• Easily transmit this dataset between stakeholders, end users and developers
What outcomes do we hope to achieve?
• Create easily deployable, update-able, and exchangeable datasets and algorithms to perform detections relevant to end users.
• Create data collection, curation, and organization processes where they don’t currently exist.
• Ensure broad access to data sets and algorithms by creating a collaborative research environment.
What resources could the lab provide?
High Performance Computers (HPC’s) and Subject Matter Experts (SME’s) in AI data collection and curation; algorithm writing; cloud resources and hosting data; Cooperative Research and Development Agreement(CRADA); Memorandum of Understanding (MOU); and other business processresources.
What would success look like?
• Datasets and accompanying algorithms/models should be able to be modified/expanded upon or tailored to specific use cases.
• Decrease natural disaster rates and associated response time, casualty rates, resources spend rates, and overall economic loss.
• Establish collaborative environments for publication of data sets and algorithms.
• Establish partnerships with entities that can provide data sets so that algorithms can be produced.
• Reducing development cost load by sharing of dataset information and technical approaches for algorithm development.
What types of solutions would we expect?
Data pipeline creation; data structuring development; create algorithm and test environment that can detect changes in building structures and construction sites; agricultural assets; or forestry, shoreline and associated overhead climatic imagery.
What's in it for industry?
Unorganized data in creates gaps of inefficiency. The aim is to create resources to solve these data problems, likely reducing costs by furthering Automation, Augmentation,and AI/ML (AAA). Create a new customer bases and support other organizations such as the Department of Homeland Security (DHS) Federal Emergency Management Agency (FEMA), and others working disaster response. Open up a wider commercial market. Results from this effort have dual use potential for both commercial and military applications. By understanding changes on resource distribution, changes in natural or man-made physical environments, transportation networks, city growth, etc., organizations can focus their efforts to developing proactive solutions to mitigate negative effects rather than reacting.
Some examples of target sectors:
• Manufacturing (Defect Sensing)
• Government (Disaster recovery)
• Insurance (Assessment)
• First responders (Local/Regional)
• Utility Companies (Grid Response)
• Media (Video/Image Processing)
The Request for Partnership Submission Is Now Over