Sourcing and Annotating Vehicle Damage Images for Automated Insurance Claim Validation

Disclaimer: Aya Data holds to the confidentiality agreements made with our clients. We shall not disclose any confidential information regarding clients and/or projects, pursuant to the individual agreements made. Consequently, client and project names and any identifiable information may be kept anonymous in our Case Studies.


The Client is a pan-African insurance company that wishes to utilize cutting-edge tech solutions to verify insurance claims. They assigned their in-house data science to develop a computer vision model that would detect and identify damage on vehicles.

The CV model would need to be able to verify damage on cars, vans, motorbikes, and buses by using images captured on smartphone cameras.

Data Acquisition

Data Annotation




London, UK

Company Size



The Client hired Aya Data to collect and annotate 85,000 images of damaged vehicles in West Africa. The Client set out strict requirements regarding how the images must be sources and how they should be classified according to the different class types of damage and the different vehicles.

After the complex image acquisition began, the images would need to be quickly annotated and quality tested before being safely transferred to the Client.

Contact us


Aya Data employed our experienced on-the-ground data collection teams in Ghana and Sierra Leone to collect the required images. After the images were sourced and curated, they were partially annotated with high-accuracy polygons by our data annotation specialists in Ghana.

The next step was to pre-label the images by developing and training a bespoke model. The pre-labeling greatly increased the efficiency of labeling the entire data set.


The Client was able to develop a highly accurate CV model for vehicle damage detection and classification as a result of Aya Data’s involvement.  The Client’s insurance claims teams were able to quickly identify whether a customer’s insurance claim accurately matched the type of vehicle listed and the degree of damage.

Additionally, the Client was able to deploy the model three weeks ahead of schedule due to the pre-labeling solution provided by Aya Data. The Client has stated that they will continue to collaborate with us on future collection and annotation projects to refine their CV models for different situations and contexts.