French scientists in Zimbabwe use facial recognition technology to identify giraffes
A team of French scientists from Hwange National Park in Zimbabwe have developed a deep learning computer system to distinguish individual giraffes, a tool that could help save this iconic African species from extinction.
Giraffes’ unique coat patterns, consisting of brown spots on a beige background, make them hard to tell apart. And researchers desperately need to distinguish between giraffes, as populations in this remote part of northwestern Zimbabwe are in decline, for reasons still being investigated.
“To our knowledge, this is the first attempt to use deep learning techniques for this task,” said lead author of the new study, Vincent Miele of the Biometrics and Evolutionary Biology Laboratory of the United States. ‘University of Lyon at RFI.
Between 2014 and 2018, his team photographed around 400 giraffes in Hwange – the park where Cecil the Lion once lived. Out of a set of nearly 4,000 images, a training dataset was created by cropping the images to show the flanks of the animals. These were fed into a computer system, known as a convolutional neural network (CNN) which analyzes visual imagery.
This deep learning system, primarily used in facial recognition of humans, was trained by scientists to re-identify individual giraffes in Hwange with 90% accuracy.
“The advantage of deep learning is that once the computer has been trained, it is very fast and can process dozens of images in a matter of seconds,” explains Miele. “Deep learning algorithms are known to outperform all other algorithms in terms of performance prediction.”
Training computers to sift photographs is essential to the work of field biologists.
Digital cameras and camera traps (remote devices that use sensors to photograph passing animals) can generate tens of thousands of images. These can get overwhelming, the authors note in a new study published in the journal Methods in Ecology and Evolution.
The team managed to train their computer program with only five photographs per animal. These images were then altered or “augmented” in the lab to create variability, making the program more effective at remembering individuals seen in the park.
Although the system is sometimes marred with poor quality images, the team observed very few incorrect matches.
“We observed very weak or erroneous matches,” Christophe Bonenfant, co-author of the study, told RFI. “When a game is reported, the result is really good and reliable.
He said the system is used to obtain important data on the group composition, life stories or movements of the giraffes.
“Like most giraffe populations around the world, the abundance is declining in Hwange and obviously everyone is looking for an explanation,” he added.
The CNN system tested at Hwange builds on previous work that uses artificial intelligence to distinguish giraffes. Having tools like these to keep an eye on animals will prove invaluable to their conservation.
Giraffes have suffered worrying population declines without the front page attention of other megafauna, such as lions, elephants and rhinos.
The International Union for the Conservation of Nature recognizes nine subspecies of giraffes. Most are classified as Near Threatened, Vulnerable, Endangered or Critically Endangered.
The Giraffe Conservation Foundation, an independent group working in 16 African countries, estimates that the continent-wide animal population has declined by almost 30 percent since the 1980s, from more than 155,000 to around 117,000 today. ‘hui.
Although the CNN system used in Zimbabwe was designed to recognize and monitor giraffes, it uses freely available software and researchers can modify it to apply to a range of other mammals. In Hwange, this could include the zebra and the kudu, a large antelope that has majestically twirled horns and unique stripes on its coat.
Bonenfant said, “We hope our system will be adopted in other sites or species.”