US 12,112,383 B2
Methods for determining image content when generating a property loss claim through predictive analytics
Jonathan Navarrete, San Diego, CA (US); Olaf Wied, San Diego, CA (US); and Norman E. Tyrrell, San Diego, CA (US)
Assigned to Mitchell International, Inc., San Diego, CA (US)
Filed by Mitchell International, Inc., San Diego, CA (US)
Filed on Sep. 24, 2020, as Appl. No. 17/031,728.
Claims priority of provisional application 62/904,976, filed on Sep. 24, 2019.
Prior Publication US 2021/0090180 A1, Mar. 25, 2021
Int. Cl. G06Q 40/08 (2012.01); G06F 16/583 (2019.01); G06F 18/21 (2023.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06V 20/20 (2022.01)
CPC G06Q 40/08 (2013.01) [G06F 16/583 (2019.01); G06F 18/2178 (2023.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06V 20/20 (2022.01)] 12 Claims
OG exemplary drawing
 
1. A method comprising:
obtaining, by a computing device, electronic images of a damaged vehicle associated with an electronic loss claim;
obtaining, by the computing device, electronic images of previously-identified damaged vehicles, wherein the electronic images of previously-identified damaged vehicles are obtained from one or more image databases, and wherein the electronic images of previously-identified damaged vehicles have been identified as depicting damaged vehicles involved in collision events;
generating, by a metadata machine learning model of the computing device, a first training data set comprising the electronic images of previously-identified damaged vehicles;
training, by the metadata machine learning model of the computing device, a first machine learning model using the first training data set;
executing, by the computing device, the first machine learning model on the electronic image of the damaged vehicle associated with the electronic loss claim to generate metadata of image features present in the electronic image of the damaged vehicle, wherein the generated metadata of image features includes a same type of feature as the electronic images of previously-identified damaged vehicles, and wherein the same type of feature is a vehicle component;
using the metadata machine learning model of the computing device, applying a Bayesian-type statistical analysis to determine a damage indicator associated with the component identified in the electronic images of the damaged vehicle associated with the electronic loss claim, wherein the damage indicator is associated with a percentage probability that the vehicle component is damaged;
updating the generated metadata with the damage indicator determined by the Bayesian-type statistical analysis;
using an image analysis machine learning model of the computing device at a first time, training a second machine learning model using the generated metadata;
executing, by the computing device, the second machine learning model to identify a subset of the electronic images associated with the electronic loss claim based on the percentage probabilities that the vehicle component is damaged;
providing, by the computing device, the subset of electronic images with the generated metadata to a client computing device to assess damage to the damaged vehicle for the electronic loss claim;
obtaining, by the computing device, feedback data on the identified subset of the electronic images from the client computing device; and
using the image analysis machine learning model of the computing device at a second time after the first time, training and refining the second machine learning model using the feedback data.