Description
As a data focused company we analysed customer’s data and suggested the use of AI for vehicle part damage identification and to combine it with Machine Learning for repair estimation and fraud detection in body shop repair estimates.

Challenge
The client had a vast set of non classified images regarding different aspects of vehicles to be repaired, the type of image and where the damage had to be found by the resulting algorithm priorly to identify the type and level of damage and the affected part.

Solution
Creation of a pipeline to process polygon data from diverse sets of classified photos, creation of a user-friendly pipeline for image damage identification for the training dataset and reduction of the number of needed images for training thanks to analysing different algorithms and use of transfer learning.

Impact
Accurate identification of damaged parts and their level of damage added a streamlined process for the estimation of repair costs combined with the Machine Learning algorithm that leverage the client’s repair historic database. Overall shortened the time to accept the repair estimate as more confidence was put in the estimates and the repair process, as well as reducing the cost of employing higher number of repair experts and their trips to the body shops.

OKORE Tech S.L.

AI, BI, data consulting, predictive maintenance

July, 2019