In Ukraine, where more than 130,000 residential buildings were destroyed and more than half a million Ukrainians lost their homes. To show the global scale of destruction caused by Russia’s military invasion of Ukraine, the UADamage project created an interactive map of destroyed and damaged real estate with the help of artificial intelligence.

The UADamage project uses artificial intelligence to compare satellite images before and after an invasion and color one of four levels of destruction.

Red dots show completely destroyed buildings. Each red dot shows lives, plans and hopes destroyed by the invasion.

Even preliminary estimates of the scale of destruction are impressive: the Russians destroyed and damaged more than 130,000 residential buildings. More than half a million Ukrainians lost their homes.

“The accounting of destroyed and damaged property will show the world what damages Russia has caused to Ukraine, and will become the basis for calculating the amount of compensation that the occupiers will pay to Ukrainians after the defeat in the war,” writes epravda.com.

UADamage was founded by a 35-year-old entrepreneur from Kovel, Vitaly Lopushansky. In January, together with his friend Oleksandr Kornienko, he decided to create NeuroMarket, a platform where you can buy an artificial neural network (a self-learning computational algorithm) to solve business problems.

How the project began

On the second day of the war, Lopushanskyi sent his wife and child to Bulgaria and signed up for the Teroboron, engaged in air reconnaissance and volunteering.

“After three weeks, I realized that we need to use our experience for the good of Ukraine. We started brainstorming, collecting successful business cases. I found an example of the use of a neural network by an American company to collect statistics on damage to houses after a hurricane,” the entrepreneur recalls.

Vitaly had heard about the terrible scale of the destruction near Mariupol and knew that no one in Ukraine uses neural networks to analyze the damage. He decided to use his knowledge to create an interactive map of the destroyed objects.

How UADamage works

One part of the neural network highlights the house in the image, and the other compares its condition with the pre-war condition. The network defines four types of damage: completely destroyed (collapse of more than 50% of the surface), significant damage (roof and wall collapse), minor damage (explosion traces), no damage detected.

Vitaly provides data on damaged buildings to the Ministry of Reintegration of Temporarily Occupied Territories and the Ministry of Infrastructure. This information can also be provided to the National Council for the Reconstruction of the Consequences of the War of Ukraine, which will conduct negotiations with countries on the reconstruction of certain regions.

In the private sector, the entrepreneur sees prospects in the sale of analytical data to insurance companies.

Problems of creating an interactive map of destruction in Ukraine

During the work on the UADamage project, the developers faced the problem of obtaining data, that is, satellite images of destroyed cities.

“Finding the data turned out to be very difficult, because it is only given to the military or high-ranking officials. It was very difficult to reach high-ranking officials: I add someone on Facebook, write, call, some assistants…” – says the founder of UADamage.

After a long search for the source of the satellite images, Vitaly met Oleksiy Damanskyi, the operations director of the Ukrainian TVIS company.

TVIS buys and sells private satellite photos, topographic maps, orthophoto plans (a terrain plan created from aerial photographs).

An archival satellite image of one square kilometer of territory costs about $20, a new image of the same territory costs $30. The area of ​​Mariupol is about 200 square kilometers, so its satellite images cost 4 thousand dollars.

“There are only a few bridges, and we wanted scale. We understood: millions of dollars are needed to digitize the whole of Ukraine,” says Lopushanskyi.