KOTA KINABALU: Artificial Intelligence (AI) may be predicted to bring the demise of humankind, but for one species at least, AI seems to be a good thing.
Conservators, experts and like-minded advocates are looking into AI technology to be used in the preservation of orangutans in Sabah.
Amanda Amran, who is part of the team, is a Computer Science PhD student from Universiti Malaysia Sabah (UMS).
She says if AI technology can take on the images from drones, and analyse orangutan nests from the drone images, then time, labour and cost implications of nest count work could be reduced significantly.
In conjunction with the International Orangutan Day on August 19, the population of the iconic species is reportedly stable in Sabah, at around 11,000 individuals, based on a 2019 study by World Wildlife Foundation (WWF).
An overall decline globally however calls for all the help the species can get.
To help scientists monitor orangutan populations quickly and accurately, WWF-Malaysia and UMS have collaborated to develop an AI technology that can automate data collection, conduct analysis and interpret drone images of orangutan nests.
Amanda's research marks the first step in exploring the world of AI for orangutan conservation.
The ultimate goal of her research is to develop a deep-learning model that will not only automatically detect and classify the nests from aerial images, but will also be able to analyse them.
This presents an invaluable opportunity for orangutan conservation as the analysis will help scientists understand the characteristics of the nests including how they're constructed and how they differ between individuals or populations.
"From aerial images, orangutan nests might look similar to other animal nests such as that of giant squirrels and eagles. Other objects that may look similar are dead trees and clumps of branches.
"On AI, just like our orangutan experts, computers must also carefully learn the features of orangutan nests. AI needs to learn in great detail nest features such as nest structure, materials, and the position of the nest on trees.
"At this early stage, we are developing a machine learning model to recognise first the general features of the nests. After that, it has to learn more intricate details," she explained.
"We can also examine how the nests are used over time, how long they last and how they change with weather patterns. By identifying common patterns in nest construction, we can better understand orangutan behaviour and their relationship with their habitat," she said.
In Sabah, despite a global decline in orangutan numbers where nearly 150,000 Bornean orangutans were deemed lost over a 16-year period between 1999 and 2015, the state has managed to safeguard its population and kept its orangutan numbers stable.
The latest estimate of 11,000 individuals matches the population estimate of a state-wide survey conducted 15 years earlier, an indication that the orangutan population in Sabah is stable, likely made possible through forest management practices and a heightened awareness of orangutan conservation.
Donna Simon, Orangutan Conservation Manager for the Sabah Landscapes Programme in WWF-Malaysia said technological innovations like AI is no doubt a crucial step towards enhancing not only orangutan conservation but also conservation in general.
WWF-Malaysia is hopeful that with innovations such as these, conservation work in this part of the world can become an exemplary model.
"We are looking at a future where technology is becoming all-encompassing and enhances almost every aspect of our lives. So why not our conservation work too?" said Donna.
She said safeguarding the orangutan population is a long-term journey that is both physically challenging and costly.
"It is very rare that we hear of orangutans being hunted and traded in Sabah. This is a testament to our collective awareness of the importance of protecting orangutans here. In turn, our awareness has significantly contributed to keeping our orangutan population safe."
One crucial part of orangutan conservation work is to monitor orangutans by mapping their distribution and estimating their population.
Scientists map and estimate the number of orangutans in an area by counting their nests. As orangutans make nests daily, finding nests in an area is a sure indication of orangutan presence.
The current traditional methods that WWF employs to count nests are ground nest surveys and aerial nest surveys via helicopters.
The former method is time and labour-intensive while the latter is cost intensive. Drones can also be employed to count nests and are thought to help reduce both cost and labour.
However, the post-processing work for drone images can take a long time. While drones can take photos of nests, it does not have the ability to accurately identify orangutan nests and count them. This work will still need to be done manually by an expert skilled at identifying nests.
For years, orangutan experts have resigned to the fact that these are the only tried and true methods that they can use in their work with orangutans. With AI, this could help improve the effort.