Drone technology seems to be exploding in two different directions: camera drones for aerial videography, and drones with AI deep learning capabilities for a wide range of different commercial and industrial purposes. The ability to get up high, take in a huge amount of visual and spatial information, and then have a GPU crunch it and do a lot of the work identifying objects, classifying them and deciding what to do about them holds great potential.
One application we didn't see coming is this, the Shark Spotter, a new initiative being tested by the Ripper Group out of New South Wales, Australia. In conjunction with local surf life saving services, the Ripper Group has been using a range of fairly serious-grade UAVs to assist with lifeguard duties.
Up to this point, the Little Ripper drones have spent a lot of their time on surveillance duties, looking for people in distress, either in the surf or on the beach. They've also been used to drop off emergency supplies like inflatable lifesavers, anchors, whistles and electromagnetic shark repellent devices.
From the shore, or even when you're out there floating in the water, sharks are more or less invisible underwater, unless they decide to go the full Jaws treatment and poke their top fins out of the water. The inability to really see what's happening underneath you is one of the reasons so many people have thalassophobia, or an intense fear of the sea.
But sharks are often very visible from directly above, so the Ripper Group is developing systems and algorithms that will give airborne drones the ability to constantly and automatically scan for sharks around surf beaches.
Training the devices begins by taking a bunch of drone footage from publicly available sources like YouTube, and manually teaching the computers what common objects look like – swimmers, surfboards, boats, sharks, waves and rocks, for example.
The computers will then start taking real-time video feeds from UAVs in the air, and start analyzing them looking for objects that meet the criteria for sharks, and generating alerts that lifesavers can then act on. This will help the developers iron out problems and increase the accuracy of the system.
In the final stage, the drones themselves will start packing their own inbuilt GPUs, probably along the line of NVIDIA's Jetson modules, and do all the processing in real time as they fly.
Once a shark is spotted, the drones have built-in loudspeakers that lifeguards can use to tell surfers and swimmers to get the hell outta there, and there's also the option of dropping shark repellent kits and other aids, as well as immediately being able to see where a shark strike occurs and get assistance to a bitten swimmer as soon as possible.
It's interesting to note that shark attacks are actually pretty rare. According to the Australian and International Shark Attack Files, there have been less than 500 fatal shark attacks since 1958, and less than 3,000 total attacks in the same period. This is most certainly one of those things where you're more likely to die falling out of bed.
But Australia and America, as well as South Africa, are most definitely overrepresented in shark death statistics, and at the end of the day, a system like the Shark Spotter drones could go a long way towards making people feel safer venturing out past the breakers. Even though the raw numbers of shark attacks are very low, the sheer terror these apex predators of the ocean can cause is very real.