Artificial intelligence boosts conservation in the Congo Basin

Google is providing their knowledge and skills in artificial intelligence (AI) in working with the Zoological Society of London (ZSL) to better understand the movements of poachers which will help conservationists to protect animals in Cameroon’s Dja Faunal Reserve in the Congo Basin.

Dja Faunal Reserve location
Two useful tags. Click either to see the articles: Speciesism - 'them and us' | Cruelty - always shameful

Dja Faunal Reserve location. Map in public domain.

The objective was to work out with more precision where poachers were operating. This would allow conservationists to better protect animals in reserves which are vast in size and where monitoring is incredibly difficult.

To achieve this goal, rather than using camera traps which have a relatively very limited range, they used devices which pick up the sounds around them up to 1 km away. ZSL placed 69 of these audio devices in the Dja Faunal Reserve. They listened out for the sounds of the animal species in reserve such as African forest elephants and weatern lowland gorillas. The devices also detected the sound of gunshots. They were placed in the reserve for a month, creating 267 days of continuous sounds.

It would have taken a very long time for people to sift through the sounds to pick out gunshots and where they occurred. This is where Google came in with their specialist skills on AI. They have a tool called Yamnet, which was trained to detect the sounds of the reserve using ZSL’s catalogue of sounds. The tool listened to the entire 267 days of recordings in 15 minutes and identified 1,746 instances of gunshots as well as the location and the time they occurred.

Acoustic device used in New Zealand by ZSL
Two useful tags. Click either to see the articles: Speciesism - 'them and us' | Cruelty - always shameful

Acoustic device used in New Zealand by ZSL. Photo: Simon Collins.

They then cross-reference the data with known instances of poaching that took place during the period of the study, at three locations. They were able to detect “poacher hotspots”, which will allow them to better prevent poaching in the future by feeding the information back to park rangers.

Now that we have a rapid, efficient way to process acoustic information, the next step is to deploy the technology at a scale throughout protected areas and to feed results back to park staff and ranges – Anthony Dancer, conservation technology lead at ZSL.

Notes: Google’s AI experts work in the cloud computing division of the company. ZSL own the London Zoo and Whipsnade Zoo in the UK. Poaching often takes place at night to evade the ranges. The recordings reveal that it sometimes take place during daytime. The reserve is a UNESCO world Heritage site.

You may also like...