Flood CamML is an open source project funded by the NSF Coastlines and People program, and was completed over ~72 hours by a group of awesome scientists from across the country.
Develop a machine learning (ML) algorithm that can detect from a single image (NCDOT Camera Feed) whether or not a roadway is flooded.
As scientists, we are interested in how often coastal roadways -- and the people that depend on these roadways -- are impacted by shallow (nuisance) flooding or ponding.
It is easy for a human to look at a traffic camera and recognize whether a roadway is flooded, but who has all day to look at web cameras?
- Can we train a machine to detect flooding?
- Or can the machine train itself to detect flooding given enough images?
- Can we create an avenue for citizen science participation to foster community engagement with science and improve model predictions?