Lodi Valley News.com

Complete News World

New Brazilian AI recognizes wild animals on the road to avoid collisions

New Brazilian AI recognizes wild animals on the road to avoid collisions

Just as GPS alerts drivers when traffic jams or accidents occur on the road, new artificial intelligence will allow warnings to be issued about wild animals on the road, preventing collisions with animals.

The new artificial intelligence system, which detects the presence of animals on the road, has been developed with research backing it Fabisp And published in the magazine Scientific reports.

Warn! Animal on the road

  • A group of researchers has developed a computer vision model based on artificial intelligence that can detect Brazilian animals, which are often victims of road accidents.
  • To do this, they trained the model with a database containing 1,823 images of the most vulnerable animals, such as the anteater, wolf, and tapir.
  • The idea is that the new system uses highway cameras to detect animals quickly and effectively.
  • It is currently being tested using different versions of the YOLO technology used for object detection.
  • In the future, researchers hope to integrate the system with GPS to alert animals on the road in real time.
Categories of animals included in the system – Image: Scientific Reports

Read more:

YOLO system

YOLO (You Only Look Once) system is a computer vision technique commonly used to identify objects. Different versions of its structure are being used in experiments on the new wildlife detection model, as it has some important advantages for the project.

One of them is single-stage detection, which allows real-time animal identification. Moreover, it can be used in mobile devices with processing power that perform relatively complex tasks.

In the YOLO experiment, videos of animals living in the São Carlos Ecological Park were used. Older versions of the system showed promising results, with animal detection accuracy reaching 80%.

See also  An ally or a villain in gaining muscle mass?

Optimization and practical tests

Although promising, the AI ​​model still needs to undergo improvements in detection in dark environments, with rain and with partially hidden animals. These issues should be the focus of researchers' attention in the next steps of the study. The team is currently seeking to establish partnerships with dealers to conduct tests in real-world situations.