The AI That Detects Shipwrecks From The Sky

By Stephen M

Thanks to this new technology, the era seems to be over of sunken ships and other underwater archeological artifacts were challenging to detect. A machines learning project undertaken in collaboration with the Underwater Archaeology Branch of the US Navy trained a computer to identify shipwrecks on the ocean floor. This becomes possible after scans by ships or aircraft on the water surface. The accuracy of the computer model created is around 92%.

Image: Shutterstock

The teaching process

The project initially focused on the coasts of the United States and Puerto Rico. Now it is ready to work anywhere in finding shipwrecks. To create this shipwreck-detecting model, the team taught the computer how a shipwreck looks. Shipwrecks can sometimes look like some seabed topographies, and as such, the computer was also introduced to the potentially subtle differences between the two. To achieve this, the researchers had to teach the computer what a natural seafloor looks like. Aside from that, developers fed it several examples of shipwrecks.

The process became easier as the National Oceanic and Atmospheric Administration has a public database of different images of shipwrecks. The images collected across the world included lidar and sonar imagery of the seabed.

Purpose of this AI

Undoubtedly, shipwrecks have something to do with human development. Finding them can help us understand past activities of humans, including trade, migration, and wars. However, this field seems not to be explored extensively due to cost and associated dangers. This is why the project comes in handy. The model reduces the time and money needed for divers or underwater drones to detect such wrecks. This is why the Underwater Archaeology Branch of the US Navy collaborated in this research. It can help the unit search for unmapped naval wrecks.

Image: Shutterstock

The project is a maiden one in the underwater archaeology field. If expanded, we can use this model to detect other sunken archaeological features like status, buildings, and even airplanes.