Universal car damage determination with make/model invariance
The present invention relates to verification of damage to vehicles. More particularly, the present invention relates to a universal approach to automated generation of a damage estimate to a vehicle using images of the vehicle and verification of a manually-generated damage repair proposals using the automatically generated damage estimate. Aspects and/or embodiments seek to provide a computer-implemented method of generating one or more repair estimates from one or more photos of a damaged vehicle and comparing the generated estimate(s) to one or more input repair estimates to verify the one or more input repair estimates.
1. A computer-implemented method of generating a damage classification for a vehicle, comprising:
receiving vehicle input data, the vehicle input data comprising at least a plurality of images wherein some of the images comprise image data of damage to the vehicle;
using a plurality of classifiers on the vehicle input data that are each specific to a different one of a plurality of normalized parts of the vehicle, each determining at least one classification of the damage to the vehicle and each of the plurality of images of the vehicle input data are processed by each of the plurality of classifiers, wherein each classifier is generic with respect to a make and model of the vehicle; and
outputting the determined classifications of the damage to the vehicle.
2. The computer-implemented method of claim 1 , wherein the vehicle input data further comprises claim input data wherein claim input data comprises details of one or more proposed parts and labor operations for repairing the damage to the vehicle, the method further comprising:
determining the parts and labor operations of the vehicle input data that are relevant to each of the plurality of normalized parts of the vehicle; and
wherein outputting the determined classifications comprises verifying that the classifications determined for each of the plurality of normalized parts of the vehicle corresponds to the proposed parts and labor operations for repairing the damage to the vehicle determined to be relevant for each of the plurality of normalized parts of the vehicle.
3. The computer-implemented method of claim 2 , wherein the determining the parts and labor operations of the vehicle input data that are relevant to each of the plurality of normalized parts of the vehicle comprises using computer-implemented natural language processing techniques.
4. The computer-implemented method of claim 2 , wherein at least some of the plurality of classifiers are arranged hierarchically.
5. The computer-implemented method of claim 4 , wherein the hierarchical arrangement comprises at least three classifiers arranged hierarchically.
6. The computer-implemented method of claim 4 , wherein the hierarchical arrangement further comprises one or more rules engines and/or databases.
7. The computer-implemented method of claim 1 , further comprising using domain adaption techniques for outputting a classification suitable for further geographies.
8. The computer-implemented method of claim 1 , wherein the classifiers further comprise a secondary classifier trained on specific vehicle parts that is operable to generate damage representations of the specific vehicle parts, wherein the specific vehicle parts are specific to any or any combination of a predetermined make, model and year of vehicle.
9. The computer-implemented method of claim 1 , further comprising the use of any or any combination of: a computer vision damage assessment model; a repair/replace prediction model; a labor hours prediction model; a blend prediction model; a paint rules model; a strip model; a remove and install model.
10. The computer-implemented method of claim 9 , wherein the any or any combination of: a computer vision damage assessment model; a repair/replace prediction model; a labor hours prediction model; a blend prediction model; a paint rules model; a strip model; a remove and install model are used as one or more secondary models that receive input from said one or more classifiers.
11. The computer-implemented method of claim 1 , further comprising using rules specific to a jurisdiction, market or geography, optionally in respect of the acceptable proportion of repair and replace decisions.
12. The computer-implemented method of claim 1 , further comprising using a rule engine and/or lookup table, wherein said rule engine and/or lookup table are used to determine the likelihood of an operation and/or an amount of work that needs to be carried out in respect of a vehicle part based on one or more the determined classifications.
13. The computer-implemented method of claim 1 , wherein the outputting the determined classifications further comprises determining one or more damage representations from the one or more determined classifications, wherein the one or more damage representations are for one or more of the normalized parts of the vehicle.
14. The computer-implemented method of claim 1 , further comprising determining any or any combination of a normalized point of impact and/or an object of collision; and determining if the classification of the at least some damage to the vehicle is consistent with the normalized point of impact and/or the object of collision.
15. The computer-implemented method of claim 1 , wherein the classification of the at least some damage to the vehicle is a generalizable damage representation for the vehicle.
16. The computer-implemented method of claim 1 , further comprising determining any anomalies in the vehicle input data.
17. The computer-implemented method of claim 1 , further comprising using machine learned models that are trained without make, model or year data.
18. The computer-implemented method of claim 1 , further comprising outputting one or more confidence scores.
19. The computer-implemented method of claim 1 , further comprising outputting one or more of any or any combination of: damage severity measures; damage locations; damage types; and damage descriptors.