US-12626230-B1 - Vehicle status analysis mechanism

US12626230B1US 12626230 B1US12626230 B1US 12626230B1US-12626230-B1

Abstract

There is provided a system for generating instructions for a generative model to present vehicle damage assessment. The system includes image sensors positioned at different heights and angles to capture vehicle images, a communication element for receiving images and accessing a stored model, and a processor connected to endpoint devices via a network. The processor generates instructional inputs combining damage severity classifications with vehicle images, sends these to the generative model, and obtains natural language elements for structural damage reconstitution. The system creates interactive presentations with populated template fields and classification selection elements displayed on endpoint devices. When users select different damage severity classifications in real-time, the system automatically generates new instructional inputs, obtains updated natural language elements from the generative model, and dynamically updates the presentation by replacing template fields. This enables real-time comparison of different damage severity assessments for the same vehicle damage using the same source images.

Inventors

Idan Cohen
Amir HEVER

Assignees

UVEYE LTD.

Dates

Publication Date: 20260512
Application Date: 20251001

Claims (17)

1 . A system for dynamically generating instructional inputs for feeding into a generative model for presenting damage to a vehicle, comprising: a plurality of image sensors positioned at a plurality of different heights and/or angles relative to the vehicle configured for acquiring a plurality of images of the vehicle; a data communication element configured to receive the plurality of images of the vehicle and to access a data storage device storing the generative model; at least one processor in communication with data communication element and with at least one endpoint device over a network, the at least one processor configured for executing a code for: generating a first instructional input including first a combination of a first classification of a plurality of classifications and the plurality of images of the vehicle indicating the damage, wherein the plurality of classifications denote increasing levels of severity of damage to the vehicle; sending the first instructional input to an input communication element for input into the generative model; obtaining from the generative model, a first set of a plurality of natural language elements corresponding to the first classification, the first set of natural language elements associated with a first indication for structural reconstitution of the damage corresponding to the first classification; generating a first state of a presentation of a template with a plurality of fields populated with the first set of natural language elements and including at least one interactive element for selecting a second classification of the plurality of classifications, presenting the presentation on an endpoint device accessing the at least one processor; and in real-time, detecting a selection of a second classification of the plurality of classifications made by a user interacting with the at least one interactive element of the presentation; generating a second instructional input including a second combination of the second classification of the plurality of classifications and the plurality of images of the vehicle; sending the second instructional input to the input communication element for input into the generative model; obtaining from the generative model, a second set of a plurality of natural language elements corresponding to the second classification, the second set of natural language elements associated with a second indication for structural reconstitution of the damage corresponding to the second classification, wherein the first state of the presentation corresponding to the first classification and a second state of the presentation corresponding to the second classification are for the same plurality of images of the vehicle of the same damage to the same vehicle; and dynamically updating the presentation on the endpoint device for presenting the second state of the presentation by replacing the plurality of fields of the template with the second set of natural language elements.
2 . The system of claim 1 , further comprising code for: in real-time, detecting a selection of a third classification of the plurality of classifications made by a user interacting with the at least one interactive element of the presentation; generating a third instructional input including a third combination of the third classification of the plurality of classifications and the plurality of images of the vehicle; sending the third instructional input to the input communication element for input into the generative model; obtaining from the generative model, a third set of a plurality of natural language elements corresponding to the third classification, the third set of natural language elements associated with a third indication for structural reconstitution of the damage corresponding to the third classification, wherein the first state of the presentation corresponding to the first classification and a third state of the presentation corresponding to the third classification are for the same plurality of images of the vehicle of the same damage to the same vehicle; and dynamically updating the presentation on the endpoint device for presenting the third state of the presentation by replacing the plurality of fields of the template with the third set of natural language elements.
3 . The system of claim 1 , wherein the natural language elements comprise human-readable text.
4 . The system of claim 1 , wherein the presentation comprises a graphical user interface (GUI), wherein the at least one interactive elements comprises at least one interactive graphical element.
5 . The system of claim 1 , further comprising code for accessing a text description of the damage, wherein the text description is included in the first combination used to generate the first instructional input and the second combination used to generate the second instructional input.
6 . The system of claim 1 , wherein the generative model is trained on a plurality of records, wherein a record is for a sample vehicle with sample damage which has been repaired, the record includes a certain classification from the plurality of classifications, the plurality of images for the sample vehicle depict the sample damage, and a ground truth including the set of the plurality of natural language elements obtained after the damage has been repaired and a total cost for performing the repair.
7 . The system of claim 1 , wherein the generative model generates a range indicating an estimated cost for structural reconstitution of the damage corresponding to a respective classification, the range corresponding to the set of the plurality of natural language elements corresponding to the respective classification, and wherein the presentation comprises a GUI.
8 . The system of claim 7 , further comprising code for: in response to interaction of a user with an interactive element of the GUI indicating a value within the range of the estimate cost for structural reconstitution of the damage, feeding into the generative model an instructional input indicating the value within the range and a request to update the set of the plurality of natural language elements of the presented GUI state to comply with the value; obtaining an update of the set; and updating the GUI for presenting the update of the set.
9 . The system of claim 7 , further comprising code for: in response to interaction of a user with an interactive graphical element of the GUI indicating a value within the range of the estimate cost for repairing the damage, generating an adapted at least one natural text element by adapting at least one of the following natural language elements of the set of the state of the GUI presented on the display: (i) at least one material used for structural reconstitution of the damage, (ii) at least one step in a repair process for structural reconstitution of the damage, (iii) labor for structural reconstitution of the damage, (iv) projected quality of the structural reconstitution of the damage, wherein an adapted set including the adapted at least one natural language text element complies with the value within the range, wherein the GUI is updated for presenting the adapted at least one natural language element.
10 . The system of claim 1 , wherein the GUI further includes an adaptable interactive graphical element indicating an estimated cost for structural reconstitution of the damage, and further comprising: automatically adapting the interactive graphical element for indicating an estimate for structural reconstitution of the damage generated by the generative model.
11 . The system of claim 1 , further comprising code for: receiving via the GUI, a value or range representing an estimated cost for structural reconstitution of the damage, the value or range obtained in response to a user interacting with an interactive graphical element presented within the GUI, feeding an instructional input into the generative model indicating the value or range, wherein the set obtained from the generative model corresponds to the value or range, wherein the estimated cost for structural reconstitution of the damage presented within the GUI includes the value or range.
12 . The system of claim 1 , wherein the plurality of natural language elements include a plurality of ranges for a plurality of values representing a breakdown of an estimate for structural reconstitution of the damage, and at least one of: the estimate is computed according to the plurality of range, and a range of the estimate is narrowed in response to selection of a specific value within a range.
13 . The system of claim 1 , wherein the plurality of fields in the template which are populated by natural language elements are selected from: steps in a repair process, cost per part, cost of labor, cost of materials, miscellaneous costs, safety considerations.
14 . The system of claim 1 , wherein at least the first instructional input fed into the generative model includes at least one vehicle attribute, wherein a record of a sample vehicle used to train the generative model includes at least one vehicle attribute of the sample vehicle.
15 . The system of claim 14 , wherein the at least one vehicle attribute is selected from manufacturer, model, year of manufacturing, installed vehicle features, and upgrades.
16 . A computer implemented method of dynamically generating instructional inputs for feeding into a generative model for presenting damage to a vehicle, comprising: using at least one processor executing a code for: receiving a plurality of images of the vehicle acquired by a plurality of image sensors positioned at a plurality of different heights and/or angles relative to the vehicle; accessing a data storage device storing the generative model; generating a first instructional input including first a combination of a first classification of a plurality of classifications and the plurality of images of the vehicle indicating the damage, wherein the plurality of classifications denote increasing levels of severity of damage to the vehicle; sending the first instructional input to an input communication element for input into the generative model; obtaining from the generative model, a first set of a plurality of natural language elements corresponding to the first classification, the first set of natural language elements associated with a first indication for structural reconstitution of the damage corresponding to the first classification; generating a first state of a presentation of a template with a plurality of fields populated with the first set of natural language elements and including at least one interactive element for selecting a second classification of the plurality of classifications, presenting the presentation on an endpoint device accessing the at least one processor; and in real-time, detecting a selection of a second classification of the plurality of classifications made by a user interacting with the at least one interactive element of the presentation; generating a second instructional input including a second combination of the second classification of the plurality of classifications and the plurality of images of the vehicle; sending the second instructional input to the input communication element for input into the generative model; obtaining from the generative model, a second set of a plurality of natural language elements corresponding to the second classification, the second set of natural language elements associated with a second indication for structural reconstitution of the damage corresponding to the second classification, wherein the first state of the presentation corresponding to the first classification and a second state of the presentation corresponding to the second classification are for the same plurality of images of the vehicle of the same damage to the same vehicle; and dynamically updating the presentation on the endpoint device for presenting the second state of the presentation by replacing the plurality of fields of the template with the second set of natural language elements.
17 . A non-transitory medium storing program instructions for dynamically generating instructional inputs for feeding into a generative model for presenting damage to a vehicle, which when executed by at least one processor, cause the at least one processor to: receive a plurality of images of the vehicle acquired by a plurality of image sensors positioned at a plurality of different heights and/or angles relative to the vehicle; access a data storage device storing the generative model; generate a first instructional input including first a combination of a first classification of a plurality of classifications and the plurality of images of the vehicle indicating the damage, wherein the plurality of classifications denote increasing levels of severity of damage to the vehicle; send the first instructional input to an input communication element for input into the generative model; obtain from the generative model, a first set of a plurality of natural language elements corresponding to the first classification, the first set of natural language elements associated with a first indication for structural reconstitution of the damage corresponding to the first classification; generate a first state of a presentation of a template with a plurality of fields populated with the first set of natural language elements and including at least one interactive element for selecting a second classification of the plurality of classifications, present the presentation on an endpoint device accessing the at least one processor; and in real-time, detect a selection of a second classification of the plurality of classifications made by a user interacting with the at least one interactive element of the presentation; generate a second instructional input including a second combination of the second classification of the plurality of classifications and the plurality of images of the vehicle; send the second instructional input to the input communication element for input into the generative model; obtain from the generative model, a second set of a plurality of natural language elements corresponding to the second classification, the second set of natural language elements associated with a second indication for structural reconstitution of the damage corresponding to the second classification, wherein the first state of the presentation corresponding to the first classification and a second state of the presentation corresponding to the second classification are for the same plurality of images of the vehicle of the same damage to the same vehicle; and dynamically update the presentation on the endpoint device for presenting the second state of the presentation by replacing the plurality of fields of the template with the second set of natural language elements.

Description

BACKGROUND The present invention, in some embodiments thereof, relates to graphical user interfaces and, more specifically, but not exclusively, to graphical user interfaces related to damage to a vehicle. Prior to repair damage to a vehicle, the damage is inspected by a professional, for example, a mechanic and/or appraiser. An estimate of the cost of damage is provided. The damage is repaired when the cost of damage is approved. SUMMARY According to a first aspect, a computer implemented method of generating a graphical user interface (GUI) presenting damage to a vehicle, comprising: for a plurality of media elements indicating damage to the vehicle: feeding into a generative model, a prompt relating to a combination of each one of a plurality of classification categories and the plurality of media elements indicating the damage, wherein the plurality of classification categories denote increasing levels of severity of damage to the vehicle, obtaining from the generative model, a plurality of sets of a plurality of human-readable text elements corresponding to the plurality of classification categories, each respective set of human-readable text elements associated with a respective estimate for repairing the damage corresponding to the respective classification category, generating a plurality of states of the GUI, each state of the GUI presenting a respective set of human-readable text elements and at least one interactive graphical element for selecting one of the plurality of classification categories, wherein the plurality of states of the GUI corresponding to the plurality of classification categories are for a same set of media elements indicating a same damage to a same vehicle, presenting the GUI on a display, and in response to a selection of a first classification category of the plurality of classification categories via the at least one interactive graphical element of the GUI, presenting a first state of the GUI corresponding to the first classification category. According to a second aspect, a system for generating a graphical user interface (GUI) presenting damage to a vehicle, comprising: at least one processor executing a code for: for a plurality of media elements indicating damage to the vehicle: feeding into a generative model, a prompt relating to a combination of each one of a plurality of classification categories and the plurality of media elements indicating the damage, wherein the plurality of classification categories denote increasing levels of severity of damage to the vehicle, obtaining from the generative model, a plurality of sets of a plurality of human-readable text elements corresponding to the plurality of classification categories, each respective set of human-readable text elements associated with a respective estimate for repairing the damage corresponding to the respective classification category, generating a plurality of states of the GUI, each state of the GUI presenting a respective set of human-readable text elements and at least one interactive graphical element for selecting one of the plurality of classification categories, wherein the plurality of states of the GUI corresponding to the plurality of classification categories are for a same set of media elements indicating a same damage to a same vehicle, presenting the GUI on a display, and in response to a selection of a first classification category of the plurality of classification categories via the at least one interactive graphical element of the GUI, presenting a first state of the GUI corresponding to the first classification category. According to a third aspect, a non-transitory medium storing program instructions for generating a graphical user interface (GUI) presenting damage to a vehicle, which when executed by at least one processor, cause the at least one processor to: for a plurality of media elements indicating damage to the vehicle: feed into a generative model, a prompt relating to a combination of each one of a plurality of classification categories and the plurality of media elements indicating the damage, wherein the plurality of classification categories denote increasing levels of severity of damage to the vehicle, obtain from the generative model, a plurality of sets of a plurality of human-readable text elements corresponding to the plurality of classification categories, each respective set of human-readable text elements associated with a respective estimate for repairing the damage corresponding to the respective classification category, generate a plurality of states of the GUI, each state of the GUI presenting a respective set of human-readable text elements and at least one interactive graphical element for selecting one of the plurality of classification categories, wherein the plurality of states of the GUI corresponding to the plurality of classification categories are for a same set of media elements indicating a same damage to a same vehicle, presenting the GUI on a di