【Bahamas】

Counting dugongs in the ocean can Bahamasbe a pretty difficult task.

That's why Australian marine mammal researcher Dr. Amanda Hodgson, from Murdoch University in Western Australia, is trying to see if drones can provide an easier and more accurate way to keep track of the endangered sea animal.

SEE ALSO: Pizza delivered by drone? It's been done for real in New Zealand

At the moment, aerial surveys of dugongs are still a largely low-tech exercise.

---

You May Also Like

---

---

They're conducted via plane, with four researchers on board. Two people have the job of counting the animals within a certain area out the window, while the other two people act as scribes, recording sightings in real-time.

Original image has been replaced. Credit: Mashable

It's no doubt a lot of legwork and costly, which is why Hodgson is looking to drones fitted with cameras to help collect data. But the aerial vehicles are a double-edged sword.

"With the drone we're taking photos continually along a straight line, so we end up with thousands of images," Hodgson said. So while the drone saves flying time, there's the arduous task of sifting through a glut of images.

That's where Dr Frederic Maire from the Queensland University of Technology comes in. A computer scientist with an interest in artificial intelligence (AI), Maire has developed an automated detection system which searches for dugongs in the drone's images.

Mashable Light Speed Want more out-of-this world tech, space and science stories? Sign up for Mashable's weekly Light Speed newsletter. By clicking Sign Me Up, you confirm you are 16+ and agree to our Terms of Use and Privacy Policy. Thanks for signing up!

Maire told Mashablethat his automated detection system, which uses Google's machine-learning program TensorFlow, has been trained to pick out dugongs by their size and color.

The detector's accuracy gets better as it learns from a wider set of positive and negative examples, the latter including anything in the sea that might look like a dugong, like wave crests and shadows, but isn't.

Maire's system currently has an 80 percent accuracy rate. It's still early days for the system, however, so Hodgson still gets someone to verify the system's identification of dugongs after they've been scanned.

Original image has been replaced. Credit: Mashable

"That task is way easier than looking at a whole image, at most it takes half the time now. That will become less as the detector gets better," she said.

With a few tweaks, the automatic detection system can be trained to identify other animals that need to be counted.

"Whales are the next step, because we already have the images to train the detector. It's basically going to be the same process for any animal you want to survey," Hodgson added.

For Maire, the detection software could potentially be applied to something like a drone-led monitoring system for sharks off a beach.

"If a large shark is identified, it could send an SMS or an image to lifeguards," Maire explained. "That's one of the potential applications we're looking at."

Topics Animals Artificial Intelligence

(Editor: {typename type="name"/})