Last year, a team of scientists and together with Yandex Cloud developers started working on a neural network for ecomonitoring of Baikal. A few days ago, the algorithm was officially launched. With the help of this project, it is possible to forecast the state of the lake and the impact of climate change on its ecosystem.
The neural network takes water samples from the lake for analysis, after which it determines and classifies the microorganisms contained in it. At the moment, the machine learning algorithm is able to recognize 70 forms of plankton that are often found in samples.
This innovation will allow biologists to simplify their work, since for many years they counted and determined microorganisms contained in water manually. Over time, the entire monitoring cycle is automated, and information for new studies can be obtained much faster.
Previously, biologists provided almost 50,000 sample images, of which 20,000 were used to train algorithms. Now the data from the microscopes are automatically transferred to the Yandex Cloud cloud platform. The algorithm recognizes small crustaceans, their species affiliation, then creates report cards. Neural network training is still ongoing in Yandex DataSphere, a service for the development and operation of ML algorithms. And the marking of information is carried out with the help of the “Toloka” service.
We will remind that the development involved the employees of the Research Institute of Biology of ISU, the Yandex.Cloud cloud service team, the developers of the MaritimeAI company, and the Fund for the Support of Applied Environmental Development and Research “Lake Baikal”.
As reported in the Yandex.Cloud blog, in the future the project participants plan to use this technology to monitor the state of water and other points of Baikal. They will also consistently make the technologies used in the project freely available.