Researchers build the first AI tool capable of identifying individual birds

New research shows for the first time that artificial intelligence (AI) can be used to train computers to recognize individual birds, a task that humans cannot do. The research is published in the British Ecological Society journal. Methods in ecology and evolution.

“We demonstrated that computers can consistently recognize dozens of individual birds, even though we cannot distinguish these individuals ourselves. By doing so, our study provides the means to overcome one of the biggest limitations in studying wild birds : recognize people reliably. ” “Said Dr. André Ferreira, from the Center for Functional and Evolutionary Ecology (CEFE), France, and lead author of the study.

In the study, researchers from institutes in France, Germany, Portugal, and South Africa describe the process for using AI to identify birds individually. This involves collecting thousands of tagged images of birds and then using this data to train and test AI models. This study represents the first successful attempt to do this in birds.

The researchers trained AI models to recognize images of individual birds in wild populations of big boobs and sociable weavers and a captive population of zebra finches, some of the most commonly studied birds in behavioral ecology. After training, AI models were tested with images of individuals they had not seen before and had an accuracy of over 90% for wild species and 87% for captive zebra finches.

In animal behavior studies, individual animal identification is one of the most expensive and time-consuming factors, limiting the scope of behaviors and the size of populations that researchers can study. Current methods of identification, such as attaching colored bands to bird feet, can also be stressful for animals.

These problems could be solved with AI models. Dr. André Ferreira said: “The development of methods for the automatic and non-invasive identification of completely unmarked and unhandled animals by researchers represents a breakthrough in this field of research. Ultimately, there is plenty of room to find new applications for this system and answer questions that seemed unattainable in the past. “

In order for AI models to accurately identify people, they need to be trained with thousands of tagged images. Companies like Facebook can do this for human recognition because they have access to millions of images of different people that users voluntarily tag. But, acquiring such tagged pictures of animals is difficult and has created a research bottleneck.

The researchers were able to overcome this challenge by building feeders with camera and sensor traps. Most of the birds in the study populations carried an integrated passive transponder (PIT) tag, similar to microchips implanted in cats and dogs. The antennas in the bird feeders were able to read the bird’s identity from these tags and trigger the cameras.

Being able to distinguish individual animals from each other is important for long-term monitoring of populations and to protect species from pressures such as climate change. While some species, such as leopards, have different patterns that allow humans to recognize them with the naked eye, most species require additional visual identifiers, such as color bands attached to the legs of birds, so that we can distinguish them. Even then, methods like this are extremely slow and error-prone.

AI methods like the one shown in this study use a type of deep learning known as convolutional neural networks, which are optimal for solving image classification problems. In ecology, these methods have previously been used to identify species-level animals and primates, pigs, and individual elephants. However, until now it has not been explored in smaller animals such as birds.

The authors caution that the AI ​​model can only re-identify the people shown above. “The model is capable of identifying birds from new images as long as the birds in those images are previously known to the models. This means that if the new birds join the study population, the computer will not be able to identify them.” Dr. André Ferreira said.

The appearance of individual birds can change over time, for example molting, and it is unknown how the performance of the AI ​​model will be affected. Pictures of the same bird taken months apart could be mistakenly identified as different individuals.

The authors add that both limitations can be overcome with data sets large enough to contain thousands of images of thousands of people over long periods of time, which they are currently trying to collect.