Where it once took months of tedious work, AI is now able to analyse month-long human findings in just days
By
The use of Google’s Artificial Intelligence tool SpeciesNet is significantly speeding up the process of tracking wildlife, say researchers at Washington State University (WSU).
Usually, this process would involve camera traps, which are motion-activated cameras placed in forests and other habitats. These generate enormous datasets – a single project produces hundreds of thousands or even millions of images that are traditionally human-reviewed to determine which species appears in each frame.
Enjoying this article? Check out our related reads…
Even with a team of undergraduate assistants and graduate students to verify identifications, WSU wildlife ecologist Daniel Thornton said that without AI, the process can take around six to seven months — sometimes up to a year, before analysis can begin.
A study was conducted to test how efficient a fully AI-automated system would be at processing a large quantity of camera-trap images collected in Washington, Montana’s Glacier National Park, and Guatemala’s Maya Biosphere Reserve.
They found that, for most species, models built from AI-identified images closely matched those produced by human experts. The results aligned in roughly 85–90 per cent of cases across key measures, such as where animals occur and what environmental factors influence them. The limited divergence for rare or difficult-to-identify species is groundbreaking, given the amount of labour that goes into wildlife-tracking efforts.
Thornton, who is also the lead author of the study, addressed that using AI is not to ‘replace people’. ‘The goal is to help researchers get to answers faster so they can make better decisions about managing and conserving wildlife,’ he said. Often, this is a core concern in ethical AI practices.
Co-author and senior staff research scientist at Google, Dan Morris, who helped create SpeciesNet, reassured that the tool helps reach the same ‘ecological conclusions’ that would usually take months to discover.
Even when the AI made mistakes, such as misidentifying animals or missing detections, researchers would argue for its usefulness rather than its limitations, as it can rule out inconsistencies more quickly. The tool’s transformative efficiency could aid smaller or underfunded conservation groups, allowing researchers to expand monitoring efforts without being constrained by data-processing capacity.
Another benefit that other organisations such as the World Wildlife Fund (WWF) have detailed, is that the innovative platform allows NGOs, governments, and citizen scientists to upload, analyse, and share camera trap data across the conservation community. This enables a more cohesive approach at tackling species at-risk and managing threats to their safety.
How else can AI support wildlife conservation schemes?
AI can also be used to support other conservation efforts, as well as tracking.
For instance, wildlife autodetection systems can combat wildlife trafficking at key transport hubs. To stop rhino poaching in some of Kenya’s highest-priority wildlife reserves, WWF collaborated with an external stakeholder, Teledyne FLIR, to install thermal cameras. Equipped with night vision and AI that can detect human, wildlife, or vehicle movements within the coverage area at night, they prove to be a useful tool for preventing crimes against endangered species.
Currently, 11 rhino sites in Kenya are using FLIR cameras and equipment in some capacity, significantly reducing wildlife poaching.
In addition, AI is also useful for a variety of other biodiversity protection purposes. Using algorithms, AI assists with ecosystem restoration planning and can predict climate outcomes, which is useful in disaster management, such as wildfire detection.
However, it is limited in some regards. Human review is needed for many other applications of camera trapping data. For example, very rare and easily confused species are still problematic for AI detection. But the findings from the recent research suggest that in some cases, image processing no longer needs to be a major constraint on large-scale camera-trapping studies.
‘The big takeaway is that this doesn’t have to be a bottleneck anymore,’ Thornton said. ‘If we can process data faster, we can respond faster, and that’s really what matters for conservation.’
