Listening to the bush: How AI can help NZ rid its wilderness of ‘hold-out’ pest possums
AI listening devices could help find New Zealand’s last remaining possums – but only if they don’t mistake bird calls for pests. Here’s how a new fix works.
New Zealand is home to a significant problem: possums, which have wreaked havoc on native forests and wildlife. The country’s Predator Free 2050 programme aims to rid the wilderness of these invasive mammals, but the final stage of eradication is particularly challenging. Removing the majority of possums through control measures is one part of the solution, but the last few survivors can quickly rebuild a population if left undetected.
Finding these elusive animals requires an effective detection method, and a promising solution lies in microphones that record sounds overnight, combined with artificial intelligence (AI) capable of scanning thousands of hours of audio for possum calls. However, AI models trained to identify possums can often produce false alarms, falsely attributing the calls of other animals to possums.
This can be a significant problem for conservation workers, who may waste time and resources searching for animals that aren't actually present.
To tackle this issue, researchers have developed a new training approach called "cross-model confusion mapping." Instead of directly asking the AI to identify possum calls, they used BirdNET, an AI system trained to recognize over 6,000 bird species. By forcing possum calls to be classified as the bird species they most closely resemble, researchers were able to identify which bird species were most likely to be confused with possums.
These bird species, known as "hard negatives," were then added to the AI's training data as examples that the AI needed to learn were not possums.
This approach resulted in models with far fewer false alarms while maintaining high detection accuracy. When tested on completely different recordings from native New Zealand forests, the retrained models produced far fewer false alarms while still accurately detecting possums. Moreover, the approach also worked well on recordings containing bird species the model had never encountered during training.
While more field testing is needed before widespread deployment, AI could become a valuable tool in conservation efforts, helping teams spend less time chasing false alarms and more time locating real pests.
Written by urgent.news from The Conversation AU's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.