Yandex AI service detects Sosnowsky’s hogweed from satellite imagery

In future, the service will learn to identify lobed echinocystis and two goldenrod species: Canadian and giant

Yandex AI service for detecting Sosnowsky’s hogweed

Experts from Yandex School of Data Analysis, working with the Stop Hogweed movement, have created an AI service that automatically identifies stands of Sosnowsky’s hogweed in satellite imagery. The service is available free of charge to environmental organisations, national parks, researchers and farmers. According to its developers, it makes image annotation 50 times faster than manual processing.

During testing, volunteers from Stop Hogweed and the Centre for Supporting Expedition Activities found infestation sites covering a combined 421 ha across 17 regions of European Russia. Moscow and Moscow Oblast have been analysed in full. Using these findings, volunteers are carrying out hogweed-control work, including digging up plants and cutting off flower heads.

Satellite imagery analysis and invasive plant monitoring

The service allows users to upload satellite images in GeoTIFF format, after which the algorithm marks areas where hogweed is growing. Yandex computer-vision technologies are responsible for the identification process. In 2026, Yandex plans to expand the service by training it to track other invasive plants, including lobed echinocystis and two goldenrod species: Canadian and giant. The plans also include creating a single map of Sosnowsky’s hogweed infestation sites using data already collected.

Sosnowsky’s hogweed is a dangerous invasive plant that spreads rapidly and can cause burns when it comes into contact with skin.