A new study has examined how computer vision and AI can be used to identify deep-sea creatures from images captured by autonomous underwater vehicles (AUVs).
Led by the University of Plymouth, the research used images taken from an unmanned submarine equipped with a high-resolution camera. In May 2016, one of the UK’s national AUVs took more than 150,000 images from more than a kilometre below sea level during a single dive in the North Atlantic. The AUV travelled roughly 3m above the seabed at 2.2 knots, taking an image every second. Around 1,200 of these were manually analysed, identifying 110 species and 40,000 individual creatures.
Using the same set of images, the team used Google’s open-source Tensorflow AI software to teach a pre-trained Convolutional Neural Network (CNN) to identify individuals of various deep-sea species found in the AUV images. The CNN demonstrated accuracy of around 80 per cent success at identifying creatures correctly, but this rose as high as 93 per cent for specific species when enough data was available to train the algorithm.
Manual analysis can vary greatly in accuracy, ranging from 50-95 per cent, but is slow and laborious. It’s hoped that the Plymouth research, published in Marine Ecology Progress Series, can lay the ground for automated deep-sea analysis that will help shed light on the darkest recesses of the ocean.
“Autonomous vehicles are a vital tool for surveying large areas of the seabed deeper than 60m (the depth most divers can reach),” said lead author and Plymouth PhD student, Nils Piechaud. “But we are currently not able to manually analyse more than a fraction of that data. This research shows AI is a promising tool but our AI classifier would still be wrong one out of five times if it was used to identify animals in our images.
“This makes it an important step forward in dealing with the huge amounts of data being generated from the ocean floor and shows it can help speed up analysis when used for detecting some species. But we are not at the point of considering it a suitable complete replacement for humans at this stage.”
The study was conducted as part of Deep Links, a research project funded by the Natural Environment Research Council, and led by the University of Plymouth, in collaboration with Oxford University, British Geological Survey and the Joint Nature Conservation Committee.