Researchers in Australia have developed an artificial intelligence tool that can count birds from drone images far more quickly than people can. Using the computer vision tool, the team says it could speed up the process by about 85 times compared with a human doing the same work by hand, opening the way to faster monitoring of bird populations.
The tool was built by Joshua Wilson, who spent years training the AI model to accurately detect birds in drone imagery. His early work focused on South East Queensland, but he soon realised that a global problem of monitoring birds needed global input, particularly for species such as migratory shorebirds that move across borders.
That realisation led to an international effort. According to the project, there was a strong response in which 33 researchers said they were happy to contribute their data. In all, about 50,000 images of birds have been collected, including on home soil in Australia, to help train and refine the model.
The motivation behind the work is the pressure on bird populations. Experts point out that there are more than 200 bird species under threat in Australia, and that the technology could make it far easier to keep track of how those populations are faring over time.
The causes of the decline are varied. Experts say habitat destruction, invasive species and climate change have all contributed to the loss of bird numbers, putting growing pressure on already vulnerable species across the country.
Monitoring those birds is not easy, because two thirds of them are often found in remote places that are hard for researchers to reach and survey regularly. By processing large numbers of drone images automatically, the AI tool is designed to help overcome that barrier.
If the approach proves reliable across many species and locations, it could become an important part of conservation work in Australia. Faster, more frequent counts would give scientists a clearer picture of which birds are recovering and which are still in decline, helping to direct limited conservation resources where they are needed most.