AI For Orangutans

Written by:

Serge Wich and Paul Fergus, of Liverpool John Moores University

Above the dense forest canopies of Borneo and Sumatra, a silent transformation is taking place to monitor orangutans. For decades, the monitoring of the three critically endangered orangutan species has depended largely on a simple but challenging task: researchers trekking through forests that are rarely easy to walk in to scan for the platforms of branches and twigs that orangutans sleep in. They occur on wet peat swamps, hills, mountains, and rarely in flat areas. This makes progress slow, time-consuming, and costly and therefore hampers covering the whole range as often as we would like to in order to monitor population changes.

These nests are more than just sleeping spots; they are the vital signs that give us information on where orangutans occur and an estimate of how many there are. But as forest loss continues, the traditional way of counting them—by foot—has become too slow to keep pace with the urgent need for conservation.

Recently, conservationists have started looking to the skies.

At first, that meant low-flying aircraft and helicopters, in which researchers scanned the treetops for the pale tangles of branches that mark a nest. The approach showed promise, and in some places it is still used, but it is expensive, limited by availability, and comes with real risk. Drones offered a safer and far more accessible alternative: relatively low cost, no crew in harm’s way, and the ability to capture thousands of high-resolution images over forest where nests might occur. Yet the bottleneck quickly shifted. The images could be collected, but finding small nests within them—one by one—was still too slow and too costly.

A new study adds another tool to the effort: Artificial Intelligence (AI). Using a deep learning model known as YOLO v10, researchers trained a system to recognise the subtle, nest-like patterns that blend into the canopy. More than 1,500 images of orangutan nests from Sabah, Malaysia, and Sumatra, Indonesia, were used to teach the model what to look for, so that it could begin to spot nests automatically—at a scale that field teams could never match on their own.

The early results suggest this approach can make a real difference.

When tested on entirely new footage, the model reached a precision rate of 98%. In practice, that means that when the system marks a location as a nest, it is almost always correct. Ground teams can then spend less time searching and more time verifying and following up in the places that matter most.

The work is still being refined. So far, the system performs a little more consistently with multirotor drones than with long-distance fixed-wing platforms, but the direction is clear. The aim is not just faster data collection, but a way to understand where orangutans are—sooner—and how their numbers are changing as forests continue to shift.

Even so, the story is not only about new technology.

Automation is not there to replace the people who know these forests best. It is there to take on the heavy, repetitive work of processing thousands of images, so expertise can be used where it counts. Time saved on screens can become time regained in the field.

That time can go into building partnerships with local communities, working with governments to strengthen protection, and pushing for sustainable practice where industry shapes what happens at the forest edge. The goal is to move conservationists from hours of manual checking

back to the work that only people can do: the careful, long-term effort of keeping habitat standing and populations safe.

As drones and AI become part of the standard toolkit, they offer a way to measure change faster and over larger areas than before. Used alongside field knowledge and local relationships, that added reach could help ensure orangutans will continue to have forests to live in.