Beach View

Automated Shark Detection Using Drones And AI

Beach View was a fun proof of concept that we did for a company in Reunion Island early in 2019. Reunion has a really high rate of shark attacks and were desperate to try something new!

There area already plenty of drones being used for shark detection in Australia. We noticed that there was a pretty low level of sophistication in these systems. Some systems still rely fully on human operators. This gives way for human error with the user monotonously searching a screen for a shadow in the water that may never appear.  With our maturing computer vision analysis software, we decided that there must indeed be a better solution.

Re-training our software we now have a working shark spotter that can with some accuracy detect sharks in the water, spot swimmers, surfers, vessels and more. Have a look at a proof of concept below:

What this enables is computer assisted (automated) live shark detection, alerting the pilot or beach patrol team during flight when a shark is found. Plotting the GPS position of the contact and showing the image of the detected shark. The patrol team is then able to make a calculated decision if there is indeed a threat to the safety of the swimmers at the beach and take the necessary action.

Aside from the actual recognition of the sharks, there is a significant flaw in current adapted models. This is the significantly short flight times per battery and need for constant piloting. Flight time ranges significantly from 20 minutes using a consumer DJI drone to up to 40 minutes for a more commercial setup. The result of this is that all day the system is going up and down, requiring significant input from the ground team costing time and money and leaving black spots in the monitoring of the beach location.

Our solution to this is an autonomous and robust, tethered drone platform that can fly all day. Our concept is as follows:

We see our solution as a fixture to a modern and safe beach, not only alerting to sharks but also other dangers. As the software adapts and matures there will be the possibility to detect a knocked-out swimmer face down in the water or a person who is getting swept out in a rip, giving lifeguards the edge saving lives at out beaches.

This is a really exciting area for us and we are looking forward to pushing the boundaries of what is possible. We would love to connect with government bodies, local councils and lifesaving organisations to trial and adapt this system and to make beaches in Australia and around the world a safer place.