Critical New Ecologist A.I. Can catalog wildlife as accurately as humans

Although Serengeti National Park is home to some of the world's most spectacular wildlife, it is home to 225 hidden cameras, known as camera traps, which discreetly portray lions and packs of hyenas crossing the Tanzanian savannah. . Documentation of these animals and their location is essential to monitor the endangered species population, preserve biodiversity, and also search for new phenomena or even species that have not yet been discovered. And until now, it has also been a huge pain.

As you can imagine, this necessary work is incredibly laborious. Together, these cameras produced three million images, all of which were manually tagged by 30,000 citizen scientists as part of the Snapshot Serengeti project, an effort to provide ecologists with the information they need to conduct the research. But thanks to a recent breakthrough, much of this work could soon be outsourced to the AI ​​

<p class = "canvas-atom web-text Mb (1.0em Mb (0) – sm Mt (0.8em) – sm "type =" text "content =" This is according to an international team of researchers who recently developed a way for that artificial intelligence solves a major ecological challenge: How to turn the millions of Serengeti photos into usable data in a They detailed their work in a June 5 article published in the journal Proceedings of the National Academy of Sciences The first author Mohammad Sadegh Norouzzadeh says Inverse this research could save ecologists from mundane tasks to give them more time to work on conservation efforts. "data-reactid =" 20 "> According to an international team of researchers who recently developed a way for artificial intelligence to solve a major ecological challenge: how to turn the millions of photos of the Serengeti into usable data in a timely manner? They detailed their work in a June 5 article published in the journal Proceedings of the National Academy of Sciences. The first author Mohammad Sadegh Norouzzadeh tells Inverse that this research could save ecologists from ordinary tasks to give them more time to work for conservation efforts

"We can make them gain time and provide them with information quickly and accurately., "he explains. "The current process they use is very slow, so it can give them outdated information.Automatic learning can provide up-to-date information for planning conservation efforts.This is why we think that it is such a critical progress for ecology. "

It would normally take Snapshot Serengeti volunteers around two to three months to label six months of images. The Norouzzadeh system can go through the same batch of images in less than an hour with human precision. But do not think this will completely eliminate the need for citizen scientists, at least for the moment.

For starters, A.I. master the art of labeling by analyzing the three million photos that volunteers had pre-labeled, which means that we are not yet at the point where we can completely outsource the work. Norouzzadeh says that the system is able to process 99.3% of the images of camera traps, but it still takes a human touch for the remaining 0.7%.