Bird watchers armed with binoculars and field books may be excellent at spotting species, but they struggle to identify individuals in a flock.
Fortunately for them, a new AI tool can make the task difficult for them, and potentially offer new insights into bird behavior.
The scientists first trained the AI in tagged images of big wild boobs and sociable weavers, as well as captive zebra finches. They then tested the system on new images of the same birds I’d never seen before. The team says that AI correctly identified more than 90% of wild species and 87% of captive ones.
[Read: Can AI convincingly answer existential questions?]
“We demonstrated that computers can constantly recognize dozens of individual birds, even though we cannot distinguish these people ourselves,” said Dr. André Ferreira of the Center for Functional and Evolutionary Ecology (CEFE), the study’s lead author. .
“By doing so, our study provides the means to overcome one of the biggest limitations in the study of wild birds: the reliable recognition of individuals.”
Building the system
Manual bird identification methods, such as attaching colored bands to their feet to distinguish them, can be stressful for animals and time consuming for researchers.
AI tools can provide a less invasive alternative, but still need to be trained on a large number of tagged images. In the human recognition systems used by Facebook, this is not a big problem, as users voluntarily upload and tag millions of their photos every day. But animals are less likely to share their personal photos.
The researchers overcame this problem by building bird feeders with camera and sensor traps. Most of the birds they studied are equipped with electronic tags, similar to microchips inserted into dogs and cats. When the feeder sensors detected one of these tags, they activated the cameras to take a photo. Eventually, the system created a dataset large enough to train AI to detect birds in the wild.
The team admits that the tool has some important limitations. You can only re-identify people who have been shown before, and you may have difficulty detecting birds if their appearance changes, such as when they throw feathers. But they believe that the largest data sets will overcome these challenges.
Ultimately, the system’s ability to identify individual animals could improve long-term monitoring of birds and help protect them from threats such as climate change. And for hobbyist ornithologists, someday it might be another useful tool in your birdwatching kit.
Posted on Jul 27, 2020 – 1:25 PM UTC
