AI Aims to Reduce Wildlife Roadkill in Brazil, Where 16 Animals Die Per Second on Roads

In Brazil, a staggering 16 wild animals become roadkill every second, amounting to an estimated 475 million vertebrate animals dying on the country's roads each year, according to the Brazilian Center for Road Ecology (CBEE). To address this devastating problem, computer scientist Gabriel Souto Ferrante has developed an AI-based solution to alert drivers to the presence of animals on the roads and potentially save countless animal lives.

Ferrante, a master's student at the University of Sao Paulo (USP), created a database of thousands of images featuring the five most vulnerable medium and large-sized species in Brazil: the puma, giant anteater, tapir, maned wolf, and jaguarundi. He then trained an AI model to recognize these animals in real-time. The project, recently published in the journal Scientific Reports, aims to collaborate with road management companies to integrate the technology with traffic cameras and 'edge computing' devices that can relay warnings to drivers, similar to navigation apps.

Previous efforts to mitigate wildlife roadkill, such as warning signs, fauna bridges, and fences, have had limited success in addressing the scale of the problem. The app 'Urubu', created by CBEE coordinator Alex Bager and other ecologists to identify roadkill hotspots, was shut down due to lack of funding. Bager emphasizes the need for a comprehensive solution as the number of roads and vehicles continues to grow in Brazil, resulting in an increasing number of wildlife casualties.

Why this matters: Brazil, the world's most biodiverse country, faces a growing threat to its wildlife populations due to the extensive road network and increasing vehicle traffic. Reducing wildlife strikes not only protects vulnerable species but also has the potential to save human lives by preventing accidents caused by collisions with animals.

Ferrante's AI-based project, if implemented, could be a promising solution to the significant problem of wildlife roadkill in Brazil. However, the initiative requires support from road management companies and funding to become a reality. As Bager states, "It is a growing problem because more roads are being built and more vehicles are being added... We need to find a solution that can keep up."

Key Takeaways

• In Brazil, 475 million vertebrate animals die on roads annually.
• AI model trained to detect 5 vulnerable species in real-time.
• Previous efforts like warning signs, bridges had limited success.
• Reducing wildlife strikes can save human lives too.
• AI project needs support from road management, funding to succeed.