Hlavní stránka Info portálu ZČU

Their methods of automatic animal recognition are now helping scientists and conservationists in the Czech Republic and worldwide. All it takes is a photo or audio recording, and the computer recognizes what animal or specific individual it is, similar to how a human recognizes a face. This automatic identification of species and individuals is the goal of research by Lukáš Pick from the Faculty of Applied Sciences and Lukáš Adam from the Faculty of Electrical Engineering. Together with other colleagues, they are developing methods that help monitor, for example, the occurrence of endangered species in nature in a non-invasive and automated way. The experts can monitor animals without disturbing them, which can happen with traditional methods, where capturing the animal and attaching a tracking device, such as a collar, is often necessary. 

Thanks to automated monitoring, these tools can detect changes in population or threats more quickly, enabling faster intervention and more effective strategies for protecting vulnerable species. Such work also generates scientific evidence that is key to setting research priorities, communicating with the public, and securing the funding necessary to sustain long-term conservation programs. The developed tools are freely available on GitHub and HuggingFace platforms. Thanks to this, they have quickly spread around the world. Experts at the Massachusetts Institute of Technology (MIT), the University of California, Jagiellonian University in Kraków, and other institutions use the tools.

"We have created a general methodology that can be adapted to different species, environments, and data sources. It can be applied to individuals of turtles, lynxes, mountain goats, toads, and salamanders," said Lukáš Picek. Lukáš Adam added: "We are launching a sea turtle conservation project in collaboration with a non-profit organization in Kenya and several universities in England."

Computer recognition of animals or specific individuals is done by scientists "teaching" the computer. In layman's terms, they take photos of turtles from various sources (cameras), for example, a computer program (artificial intelligence) learns to recognize shapes, colors, patterns, etc. The program can then automatically find and sort photos of turtles and identify them immediately. "The technology makes it possible to find out where and when a specific turtle migrates, how often it returns, and whether it is in good condition. Thanks to information about the movement and number of turtles, non-profit organizations can better protect beaches from poachers or inappropriate tourism," explained Lukáš Picek.

Growing interest in Czech research opened the door for Lukáš Picek and Lukáš Adam's team to one of the world's most important conferences in the field of computer vision, the Computer Vision and Pattern Recognition Conference (CVPR), which took place this year in Nashville, USA. The two scientists, in collaboration with Sara Beery from MIT, organized their own professional tutorial focused on animal identity recognition. It attracted the attention of experts from prestigious universities and environmental organizations worldwide. A tutorial recording will be available on YouTube in the fall.

In addition to experts from the University of West Bohemia, MIT, the University of Massachusetts, Brigham Young University, the University of Pittsburgh, and the WildMe organization also participated in the program. Both technical challenges (such as open data and adapting models to new environments) and specific case studies, from elephants and lynxes to whales, were discussed.