Material Detection AI
Using artificial intelligence (AI), we developed a model that identifies materials in damaged buildings in Bucha, Ukraine. This model can recognize materials with up to 86% accuracy, depending on the type of material.
Project team:
Data analysis and development — Herman Mitish, Kateryna Lopatiuk
Research — Svitlana Usychenko, Roman Puchko
Communications — Natalie Shvets
Partners:
UNDP
Kyiv School of Economics (KSE)
Damaged in UA
Supported by:
UNDP
KSE
Year:
2023
Documents:
Project Inquiry
What if AI could analyze images of war-affected buildings in Ukraine, creating a detailed public database of their materials?  

What if we build a case study in a city like Bucha, so we could see how effectively AI identifies key resources like concrete and metal for reconstruction?  

A successful approach could then be replicated across affected regions, promoting a more sustainable and informed rebuilding process by maximizing salvageable materials.
About
the Project
We trained an AI model to identify materials present in damaged and destroyed buildings in the city of Bucha. Leveraging data on destruction from Damaged In UA, this AI-model achieved an accuracy of up to 86% in classifying materials, with some variation depending on the type.
Drone images of the damaged buildings in Bucha, along with data on the level of damage and from open sources, formed the basis for the analysis. The results are available on an interactive web platform featuring two neural network models. The latest versions of the trained model excel at identifying metal sheets, while brick identification remains the weakest at 37% due to limited visibility in the images.
The AI identified materials like slate, metal tiles, concrete, bituminous roofing, and metal sheets most frequently.
These materials from destroyed buildings hold significant potential for reuse or recycling. Concrete, metal tiles, and metal sheets offer a 100% recycling rate. However, slate requires 100% disposal due to the presence of asbestos, a hazardous carcinogenic material.
Project Impact
Our methodology demonstrated its effectiveness in assessing damaged buildings across Ukraine. This allows for a more precise evaluation of destruction levels, leading to a better understanding of reconstruction needs across the country.

The approach is particularly valuable in areas with limited access due to minefields, where traditional ground-based assessments are not feasible.  Ultimately, this project enhances Ukraine's ability to evaluate destruction and offers hope for sustainable and informed reconstruction efforts.