Ecotourism in Crimea aims to preserve the environment
Russian specialists are creating a system for identifying Amur tigers in frames from photo traps using artificial intelligence. The identification of rare predators is carried out according to the unique pattern of stripes on the skin. The project was initiated by the Amur Tiger Center and is being implemented jointly with Sberbank and the Far Eastern Artificial Intelligence Center of the Far Eastern Federal University. This was reported by the Wild Nature of Russia portal.
"A similar principle has been used for a long time, but with the help of a digital tool, the speed and quality of processing a large volume of data will be significantly increased," - the message says.
According to the General Director of the Amur Tiger Center, Sergei Aramilev, in the early 2000s, when automatic photo traps were just beginning to be used in Russia, they used film. These devices took few photographs, which were processed by specialists manually. In the 2010s, digital photo traps appeared en masse, which began to take not 12 or 36 pictures, but from 1000 to 3000 frames and recorded them on memory cards.
Scientists and employees of protected areas had difficulties - they had to separate photographs with tigers from images of other animals and pictures that were obtained in case of false triggering of photo traps.
"In those years, a large array of data with leopards and tigers had already accumulated, a lot of time was spent on manually sorting through this entire archive and comparing animals by the patterns on the skin," the portal notes.
The procedure sometimes took up to six months, depending on the territory where the scientists worked. Programs began to appear that were designed to facilitate the work of specialists in sorting photographs, but none of them worked properly.
Therefore, the Amur Tiger Center decided to develop a new product jointly with Sberbank, which has expertise in the field of artificial intelligence.
"We have a large database of tigers that have already been identified, and an understanding of how it should work. On the part of our partners, there are programmers and specialists in the development of artificial intelligence," - the portal quotes Aramilev.
The AI model already identifies tigers, but the system takes time to be trained. "The more data the program processes, the more it learns and shows a higher result. Now specialists are helping the system and artificial intelligence to develop, they check and correct the results of the work. The creators of the project hope that by the fall the system will be able to provide significant assistance to scientists," - the portal reported.
According to Aramilev, the promptness in processing this data is very important.
He gave an example: while sorting through photographs, specialists saw a photo of a tiger with a broken paw, which needed to be helped within a week, but this information did not reach people until six months later.
"Any conservation of species, population management requires prompt receipt of information. Otherwise, photo traps become a way of obtaining beautiful photographs, and not a tool for protection," he said.
The developed system allows processing thousands of images. Images are loaded into the program, which already independently determines whether there is a tiger in the frame, and then, reading the unique pattern, identifies a specific individual if the identified tiger is already in the database.
"If the individual has been photographed for the first time, then they immediately issue a "passport". The new development will help to keep records of the population, kinship ties, determine the range area and track the movement of individuals," - the message says.
Each tiger in the system will have its own folder, where all photos of a specific individual will be stored.
In January, the project of the Amur tiger recognition system was submitted for the World Summit on the Information Society (WSIS) award in the E-Science category. WSIS is held under the auspices of the UN. The project was included in the final selection for the award among 360 participants.
Blogs
