AI in building maintenance: shaping the future of heritage protection

Thanks for talking with us today, Lukas. Could you please start by introducing us to the project?

The main goal of this GovTech project was to create an app that streamlines the process of creating reports on cultural heritage buildings. The report includes an overview of the building, along with both automatically and manually detected defects from pictures taken by the user, their causes, possible fixing solutions, and the building’s maintenance plan.

Could you please elaborate more about the problem that required a new solution?

Currently, the building maintenance process is both labour and time-intensive. A team of analysts has to visit the building, take photos, identify defects, and then manually write a report. With our solution, the owner can now receive the report instantly. They simply need to have photos and other information about the building ready and upload it onto the app. Then, they can review the defects that our model outlines. After this checkup, a PDF report is generated and delivered to the user.

Can you elaborate on your personal view of how it will affect heritage protection?

Companies or users of this app can save time and resources in the report creation process, so this project contributes to the broader trend of process automation, where manual processes are transferred to IT systems. This is particularly relevant for public sector agencies like our client – together we can ensure that the state is equipped with the most modern solutions and ready for the Digital Age.

Do you see the potential for this product to be used outside of Lithuania? Can it be adapted elsewhere?

The app features an English language translation, which makes it adaptable for similar systems in other countries. There is no doubt that every country has a substantive architectural heritage that needs constant protection, so there are plenty of scalability options. 

What was the greatest achievement of this project?

The greatest achievement was fulfilling all the main requirements of the app within the very short time frame that was given to us. The main requirements were:

1. Building Information Input Feature, allowing the user to enter information and upload photos of the building, and was the least complex to implement.

2. Automatic Building Defects Detection Feature, enabling the user to receive automatically detected defects. It was the most complex feature as it required training a defect detection model using a large amount of the client’s data, including thousands of photos of various buildings and different types of defects.

3. Manual Building Defects Detection Feature for the user to update automatically detected defects in case the model makes a mistake. We aimed for high user-friendliness, so it had medium complexity.

4. Report Generation Feature, so that the user can download the generated report on the building information and identified defects. This feature involved building a rule-based model to handle different scenarios.

Were there any risks associated with the project, and how did you mitigate them?

I believe the greatest risk was not finishing the project on time. To mitigate it, we maintained the app’s low complexity and put the most important requirements first. We achieved lower complexity by creating a simpler system architecture. The system is stateless, and all processes are facilitated by continuous communication between the user interface and the server.

Do you still participate in the maintenance of this model, or has the company taken over?

Yes, we are continuously improving and maintaining the app. We are working to enhance the accuracy of the defect detection model and the design of the user interface. For maintenance, we address bugs based on client feedback and ensure general accessibility.

For more information about AAI Labs’ latest AI-powered solutions for public sector, see the publication here.