Hundreds of meters below the ocean’s surface, where sunlight gives way to darkness, coral reefs play host to an impressive array of bottom-dwelling species. This deep-sea reef is home to squat lobsters, sea spiders, crabs, brittle stars, and soft corals, among many others.
Corals are tough, but fragile. They are easily damaged by the nets that are towed along the sea floor by industrial fishing vessels. Once damaged, these fragile marine ecosystems are slow to recover. To help protect deep-sea coral reefs, mapping and monitoring is a top priority for scientists, conservationists and sustainable fisheries.
Monitoring the sea floor is an expensive and difficult challenge. One solution is image-based surveys. This involves towing a rig of cameras behind a ship to take high-quality images of the sea floor. This part is simple enough.
But then, the scientists must go through each image, frame by frame, to determine the features of the sea floor. They are looking to distinguish corals from things like coral debris, debris from other organisms, sand, pebbles, and rocks.
Using deep learning to help monitor the deep sea
This time-consuming and effort-intensive process creates a bottleneck. It limits the amount of data available to make decisions about strategies that work to protect deep-sea coral reefs.
This is where artificial intelligence (AI) comes in. Our researchers have developed a deep learning system that can analyze images to identify and quantify deep-sea corals, in a fraction of the time it takes a human.
The system consists of a multi-layered network of artificial neurons. It aims to simulate the human brain by learning how to identify complex patterns within data.
“The deep learning system is incredibly fast and accurate,” said Chris Jackett, one of the study’s authors and research scientist at CSIRO.
“The final trained model classified more than 2,300 images in less than 20 minutes – a task that could take more than three months,” he said.
It learned to distinguish between six features of the sea floor—anchored coral, coral rubble, debris from other organisms, sand or mud, pebbles or pebbles, and rocks—with an accuracy of 98.19%. In some cases, his performance was more consistent than that of a person.
The model was trained using images collected from our RV investigator, during a trip off the south coast of Tasmania in 2018.
Our depth camera system recorded continuous video and took pictures every five seconds, at depths ranging from 600 to 1,800 meters below the surface.
The researchers manually reviewed nearly 6,000 images, which were then used to train the deep learning system. The raw data set consists of 140,000 data points, or “snippets”.
Training AI to avoid image spoilers
As it turned out, the data was very “noisy”. A lot of the scraps contained a mixture of different features, were too dark or too light, or were bombarded by native animals such as seastars or hedgehogs.
Our researchers set to work cleaning the data, whittling it down to about 70,000 snippets that were clean enough for the model to learn from. They tried different methods of training the model, and different network structures, until they came up with a combination that achieved an accuracy of 98.18%.
When the final deep learning model was tested on unclean data it had never seen before, it performed at least as well as a person. It was consistently accurate, even for complex images with lots of features, which can be difficult for humans to classify.
Artificial intelligence is giving our researchers a helping hand to protect coral reefs
While AI can greatly speed up this process, research still needs a human touch.
“We think having a person in the loop would be the most efficient way to use this deep learning model on new images,” said Chris. “When the model is uncertain about how to classify something, a person can step in to help the model learn.”
AI comes to the rescue to protect deep sea coral reefs and other fragile ecosystems. In the future, it could be used by deep-sea fisheries – such as New Zealand’s deep-water group – to reduce impact on the environment.
It offers great potential in addressing a major conservation challenge facing our world’s oceans.
the quote: A deep learning system can analyze images to protect deep-sea coral reefs much faster than humans (2023, May 18) Retrieved May 18, 2023 from https://phys.org/news/2023-05-deep-images-deep -sea – coral reefs. html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.