Sometimes, less truly means more. Leveraging the expertise of our mathematical experts, we developed a lightweight and straightforward solution that enables remote estimation of damage size, eliminating the need for on-site visits.
A Dutch-Polish software company
Our client is a Dutch-Polish software company with over 18 years of IT development, support, and maintenance experience. They support top-tier global players, including the largest independent expert organization in the Testing, Inspection, and Certification (TIC) sector.
Digitalizing Damage Assessment
Faced with the COVID-19 pandemic, the global TIC company needed to adapt its damage assessment process for efficiency and safety.
They wanted to replace the traditional method of on-site evaluations by consultants with a mobile app, where customers could upload photos of the damage, and the system would automatically calculate the damaged surface.
This move addressed the immediate pandemic challenges and was a step in modernizing the company’s insurance claims handling.
Designing a Compact App
Our direct client, who was responsible for delivering the app, had the foundational idea of using everyday objects like a credit card or an A4 sheet as reference points to calculate the area of damage on a surface, but they lacked the specific mathematical expertise to bring this concept to life.
Our job was to design this system accurately, efficiently, and user-friendly, given that the responsibility for damage size estimation was on the user.
Additionally, considering the download limits on platforms like Google’s Play Store, the solution needed to be compact without significantly increasing the app’s size.
It was a classic case of less being more, where avoiding overengineering was essential.
Utilizing simple mathematic models
Our team was composed of two members: a doctor in mathematics specializing in algorithm development and a university professor serving as a Python developer. Rather than creating a complex AI engine, they utilized straightforward mathematical models and classical methods for problem-solving.
Their approach involved using everyday objects, such as a credit card, as reference points for calculating damage size. By recognizing the constant dimensions of these objects, we could accurately measure distances between points marked on images.
They applied perspective transformation techniques to render these images into a two-dimensional representation, thereby correcting any distortion due to perspective. This process converted the image into a flat plane, employing planimetry to determine the damaged area based on these marked points.
(And by the way – planimetry is just a sophisticated term for ‘measuring objects on a flat surface.’)
Lightweight technology
Technically speaking, the solution was designed as a streamlined microservice architecture using Fast API, Celery, and other relevant technologies. It was seamlessly integrated into the existing infrastructure, facilitating efficient communication without adding unnecessary complexity.
In addition, to achieve precise measurements, the system only required the points marked on the credit card for perspective correction, not the entire image.
This approach simplified the process and heightened privacy and data security by eliminating the need to transfer complete images.
Two Weeks to Full Functionality
Our experts delivered a fully functional system within a two-week sprint.
The application consistently provides precise analysis results in just about one second, significantly reducing the manual workload and improving customer response times.
We guaranteed that the solution would remain operational for six months after implementation, and it lived up to that promise—operating seamlessly without interruptions or downtime for an entire half-year.
While the system is lightweight and straightforward, it is robustly prepared for future development.
