- Posted by MGUG Admin
- On September 4, 2021
The research crew that had hastily assembled off the coast of Santa Catalina Island in Southern California, some 3,000 feet above the ocean floor, could hardly believe what they were seeing. They knew they would find dumped barrels resting on the seabed, they just didn’t realize how many. Years earlier scientists had detected DDT pesticides seeping into underwater sediment and found traces in the fat of fish. A targeted dive even provided alarming video evidence of barrels degrading, ringed by fluorescent traces of leaked materials.
Despite amassing more than 100 gigabytes of sonar data, the data set itself wasn’t so big it couldn’t be managed. The challenge was that there were a tremendous number of barrel-like objects in a relatively small area, Batchelor said. Researchers didn’t want to overlook anything or risk counting the same barrel twice. They also wanted to ensure that they counted and mapped just what were barrels and not other objects. They sought help from Scripps colleague Dr. Sophia Merrifield, an associate researcher who specializes in machine learning, to accurately characterize objects in imagery. And with her help their results greatly reduced anomalies.
Full Article At: