October 17, 2018 – Janet Mann, Georgetown professor of biology and psychology, is collaborating with Google’s artificial intelligence engineers on individual identification of wildlife through images.
Mann, who has studied over 1,800 individual dolphins in Shark Bay, Australia, provided Google engineers with tens of thousands of images of Shark Bay bottlenose dolphins. Her Shark Bay Dolphin Project has been in existence for 32 years.
Google developed a fin-matching system within their new Cloud AutoML (ML for Machine-Learning) that works in seconds with high accuracy, regardless of how many dolphins there are in the image or at what photograph angle.
Critical for Conservation
“Tracking individuals is a critical part of biology because natural selection acts on the variation between individuals,” Mann explains, “and it’s also important for conservation, management and appeals to the public in terms as they are interested in dolphins as individuals.”
Matching work by sight can take hours by humans, and the new system can track individual dolphins that migrate across hundreds of miles and over a decade or more.
Mann is applying the system to her latest research, The Potomac-Chesapeake Dolphin Project, which studies nearly 1000 dolphins named after significant historical figures throughout U.S. history, including former First Lady Barbara Bush and former President Jimmy Carter.
“We know the individual dolphins in Shark Bay and can recognize them on sight,” Mann says. “But how do you track animals that you know less about?”
Using Google’s system, Mann was able to match the Potomac-Chesapeake dolphins with those of the Mid-Atlantic Bottlenose Dolphin Catalog, which comprises photographs of dolphins at 18 sites in the western Atlantic and is housed at Duke University.
Using Google's AI system, Mann learned in seconds, for example, that the dolphin “George Mason” winters off the coast of North Carolina, after the system matched photographs of him in 2008 and 2010 to photos of ‘George’ in the Potomac in 2015.
The partnership with Google will allow Mann's Potomac-Chesapeake Dolphin Project, which is attempting to better understand and protect dolphins in the Potomac, to proceed at a much faster rate.
“We are hoping this collaboration with Google will create a product that can be applied to all wildlife that have individually identifiable characteristics and will help boost conservation efforts all over the world,” Mann says.