The platform uses custom image recognition algorithms to detect structural damage and provides preventive maintenance recommendations. Integrated with Google and Outlook calendars, it allows for personalized maintenance schedules, ensuring timely repairs. The system aims to prevent deterioration and minimize costly restorations. 

Context

Preserving cultural heritage is a critical challenge , with over 26,000 historic buildings in Lithuania requiring ongoing maintenance. Many of these structures are important to the country's cultural identity but are affected by  environmental factors and aging of materials. Fixus Mobilis plays a crucial role in preventive care, yet its reach is limited. With only three mobile teams visiting 70 buildings annually, some heritage sites remain vulnerable. Manual inspections often can miss early signs of damage, leading to costly restorations or irreversible loss. To address these limitations, Fixus Mobilis needed a scalable, technology-driven solution to upscale preventive maintenance efforts​.

Solution

AAI Labs developed an AI-driven platform designed to modernize the inspection and maintenance process of historic buildings. This solution uses advanced image recognition to detect structural damage, such as cracks, rot, and mold, by analyzing photos uploaded by users. Once damage is detected, the system provides detailed reports, suggesting preventive maintenance steps. Additionally, the platform integrates with Google and Outlook calendars, generating customized maintenance schedules and reminders, ensuring timely repairs and reducing the risk of further deterioration​.

Implementation stages

1. Data collection & preparation:

The initial phase involved gathering and annotating over 5,000 images of historical buildings. This dataset served as the foundation for training AI models to recognize damage across different architectural styles​.

2. AI development:

We developed convolutional neural networks to identify various forms of damage, such as plaster cracks and wood rot, from user-submitted images. Natural Language Processing solutions were employed to generate textual recommendations for addressing the detected damage​.

3. Integration & testing:

The system was integrated into the Fixus Mobilis web platform, allowing users to access the tool seamlessly. Beta testing involved user feedback to refine both the image recognition algorithms and the user interface, ensuring smooth functionality.

4. Prototyping & iteration:

Continuous improvements were made based on beta testing, including enhancing detection accuracy and providing more user-friendly maintenance plans. Calendar integration features were fine-tuned to ensure effective reminders and task scheduling​.

Developed prototype

Using the Virtual Fixus Mobilis Assistant prototype is straightforward and intuitive. Users begin by entering their building details, such as construction year and materials.