Facial recognition style artificial intelligence has been created…for birds.
It can tell individuals apart – which humans find impossible. The machine learning tool is up to 92% accurate – and will boost conservation efforts.
It may also make monitoring devices that are attached to animals redundant – which can cause stress.
The technology is similar to software that remembers faces on social media.
It is the first time a computer has been trained to recognise our feathered friends.
Companies like Facebook have access to millions of pictures of different people that are voluntarily tagged by users. But acquiring such labelled photographs of animals is difficult – and has led to a bottleneck in research.
The international team overcame this challenge by building feeders with camera traps and sensors.
Lead author Dr Andre Ferreira said: ‘We show computers can consistently recognise dozens of individual birds – even though we cannot ourselves tell them apart.
‘In doing so, our study provides the means of overcoming one of the greatest limitations in the study of wild birds – reliably recognising individuals.’
Bird populations around the world face a number of existing pressures including climate change, intensive farming and deforestation.
Global warming has already had a significant impact on numbers – increasing the extinction risk for many species
The study published in Methods in Ecology and Evolution involved collecting thousands of labelled images of birds.
They included wild great tits and sociable weavers and captive zebra finches – among the most studied animals in nature.
After programming, the neural network successfully identified over 90% of the former two – and 87% of the latter.
The deep learning technique specialises in classifying images.
It has previously worked on larger animals at a species level – as well as specific primates, pigs and elephants.
But it had never been explored in smaller creatures such as birds – until now.
Dr Ferreira, of the University of Montpelier, France, said: ‘Deep learning has the potential to revolutionise the way researchers identify individuals.
‘To our knowledge, this is the first successful attempt at performing such an individual recognition in small birds.’
In behaviour studies, individually identifying animals is one of the most expensive and time-consuming problems – limiting their scope and range.
Current methods like putting colour bands on birds’ legs sometimes cause them anxiety. These issues could be solved with AI.
Dr Ferreira said: ‘The development of methods for automatic, non-invasive identification of animals completely unmarked and unmanipulated by researchers represents a major breakthrough in this research field.
‘Ultimately, there is plenty of room to find new applications for this system and answer questions that seemed unreachable in the past.’
The birds belonged to populations in South Africa and Germany. Most carried a PIT (passive integrated transponder) tag.
This is similar to the microchips implanted in pet cats and dogs. This enabled antennae on the feeders to read their identity – and trigger the cameras.
Being able to distinguish animals from each other is important for long-term monitoring – and protecting species from climate change.
Some – like leopards – have distinct patterns that allow humans to recognise them by eye.
But most require additional visual identifiers – such as ringlets or tiny backpacks.
These methods are extremely time consuming – and can be error prone.
Dr Ferreira said: ‘Overall, our work demonstrates the feasibility of applying state-of-the-art deep learning tools for individual identification of birds, both in the laboratory and in the wild.
‘These techniques are made possible by our approaches that allow efficient collection of training data.
‘The ability to conduct individual recognition of birds without requiring external markers that can be visually identified by human observers represents a major advance over current methods.’
His researchers now plan to improve the AI model by collecting images of thousands of individuals over long periods of time.
Currently, it is only able to pick out those it has been shown before.
Dr Ferreira said: ‘It is able to identify birds from new pictures as long as the birds in those pictures are previously known.
‘This means that if new birds join the study population the computer will not be able to identify them.’
The appearance of individuals can also change over time – through moulting, for instance.
Both these problems can be overcome with large enough datasets, Dr Ferreira added.
0 Commentaires