TÜV Rheinland and Adomea have collaborated to create an automated vehicle damage assessment system capable of producing a comprehensive image of a vehicle's surface within a minute. Adomea's measurement setup has created a valuable database with untapped potential, prompting TÜV Rheinland to partner with Merantix Momentum to leverage computer vision technology to enhance the damage assessment process further.

Merantix Momentum developed a fully automated system to detect defect classes, such as scratches, abrasions, dents, etc., on the car body surface, setting new standards in vehicle damage detection.

Using a computer vision model, Merantix Momentum utilized a multi-layered database that included RGB values, curvature and reflectivity. The biggest challenge of the project came from the complexity of the data and in particular its uneven distribution - less than 1% of the vehicle image data contained damage. In addition, the damage present was minimal, making it difficult even for experts to determine whether the images actually showed damage. To solve this, Merantix Momentum introduced a new approach to annotating the data to quickly improve its quality. Working with an agile and data-centric methodology enabled the rapid generation of annotated data and resolved the imbalance in the original dataset.

Based on multimodal sensor data, Merantix Momentum has developed an innovative AI solution to automatically detect vehicle damage. This solution is customizable for a wide range of applications, including final inspection, vehicle logistics, lease return and other scenarios that require accurate damage assessment.