How does it work?

Using this application, you receive a reliable damage estimation directly! But, how does this work? It starts by taking a picture of the damaged region. After a quality check the image will go through an AI model, which will detect the damaged parts of the car together with the exact contours of the damage. The user then receives an image in which the damaged area and the damaged car part are highlighted and is asked to provide feedback. The user can correct the contours of the damage if necessary. To reach a damage severity assessment as accurate as possible, the user will receive images of similar damages, slightly varying in severeness. The user will swipe the pictures most similar to its own damage. The severity score will be adapted and finally the user will receive the results of the damage severity. And this all is done within a few minutes.  


  • Q: How does Bump-Out! differ from conventional damage claim apps
    A: Conventional damage claim apps require human work to identify damages on send images. Bump-Out! uses new innovations in AI to computationally recognize damages from images and return an estimate of the damage’s severity. This way Bump-Out! not only provides a more sophisticated user experience but also a reliable damage estimation tool which gives less exposure to human error in assessing damage, while also making the process less labor intensive, causing considerable cost reductions.
  • Q: When will the Bump-Out! app be launched?
    A: Bump-Out! will be launched at the first of January 2021.
  • Q: What happens when Bump-Out! fails to recognize a damage correctly?
    A: When the AI model only recognizes part of the damaged area, the user can draw the actual surface of the damaged area on their screen. In the rare case the AI model does not recognize the damage at all, the user is advised to contact the support service of their insurance company. In both cases, Bump-Out! will use this user-feedback to ongoingly improve its model, to provide even better services to future users.



One response

Geef een antwoord

Het e-mailadres wordt niet gepubliceerd.