Artificial Intelligence technology created at Liverpool John Moores University is being used by conservationists to identify wounded animals and plot their movements in the wild.
Professor Paul Fergus and Doctor Carl Chalmers spent three years working on this technology to help the work of conservationists.
They also had input from Professor Serge Wich and Professor Steven Longmore, also LJMU, to further develop the technology.
They work closely with Knowsley Safari Park to use their technology in a bid to help them find ways to improve the animals’ welfare.
Professor Paul Fergus told Merseynewslive: “There are already cameras used by Chester Zoo that can sense an animals’ movement and identify them.
“But, to process 1.5 million pictures would take one to two scientists about two years to assess all that data.
“Now we can process those images in the space of 24 hours using our technology.”
So how does this technology work?
Professor Fergus said: “You give the technology lots of images and then keep telling the algorithm what animal it is. So when it comes across that species, it can identify it.”
The algorithm that both Professor Fergus and Dr Chalmers developed is constantly learning and evolving.
Yet, they have successfully got it to be 95% accurate. That means with every 20 images of the animals it’ll be able to identify 19 of those correctly.
Professor Fergus said: “Our system ‘conservation AI’ can detect those animals, classify them and respond back to us within seconds. We can notify the individual and let them know about the animal we’ve just seen.”
They have just finished another project creating technology identifying individual orangutans using facial recognition.
This means that they can now tell the difference between orangutans by their facial features, which has never been done before.
Dr Chalmers said: “We do it all for free, we don’t charge for data analytics.
“We’ve even gone on to create a tropical reef model so we can monitor and identify fish in the ocean. We also work with drones and acoustics so we can analyse bird calls and animal noises.”
Why focus on conservation AI technology?
Both scientists wanted to get an understanding of the specific issues conservationists were having when trying to monitor different animals.
Speaking to different conservationists from South Africa and working with researchers in Borneo and Cambodia has helped guide their conservationist AI.
Professor Fergus said: “We both have a software development background; we’ve done work for the NHS before.
“We’ve been using machines for wound analysis in a clinical trial that’s just finished. But that same model we found can detect wounds on an animal as well.”
Developing this technology does take time and there have been some barriers that the team have come across.
Dr Chalmers said: “Obviously finding pictures of lions is a lot easier than trying to find images of other species which can be hard especially when you’re training an algorithm which is data hungry.
“Also, trying to train it to catch different angles and still work in different environments was challenging.
“So, we deployed this approach called transfer learning where we can start training baseline models for as little as 1500 images, so that’s been a big limitation in the past training these models but using this approach has been able to resolve the problem.”
They’ve already been congratulated by conservationists in South Africa as it’s made a huge impact to their day-to-day work.
They are now set to head there in May to deploy 15 cameras to help conservationists monitor more species, especially larger animals such as elephants.
The team are one of the main leaders worldwide in producing the conservationist AI technology and have only just scratched the surface of just how much more they can evolve their work.
Featured image from Sutirta Budiman/ Unsplash
