Method and system for virtual reconstruction of accidents
A system and method for generating simulations of a vehicular accidents and detecting inconsistencies between different sources of data for the accidents are disclosed. The system obtains reports and other data describing the accident and processes the information for use by a keyword model. The keyword model is configured to detect terms in the data that are more likely to provide insight into the accident. The keywords are provided to a generative engine that is configured to generate a simulation of the accident based on visual elements corresponding to the keywords. In some cases, the simulation may include an animated video and/or a 3D model of the accident. Individual simulations may be generated for each source of data obtained. The simulations can then be compared to detect potential inconsistencies.
1. A computer-implemented method of identifying inconsistencies between different sources of data for a vehicle accident, the method comprising:
receiving first data including a testimony of a witness to the accident;
receiving second data including information about the accident from a source other than the witness;
performing natural language processing (NLP) on the first data;
providing the processed first data to a keyword machine learning model to extract one or more terms detected in the first data that are associated with high degree of relevance for the accident;
feeding the extracted terms to a generative engine to generate a first simulation of the accident based on elements represented by the extracted terms, wherein the first simulation includes a first animated video depicting a first reconstruction of the accident;
generating a second simulation of the accident based on information extracted from the second data, the second simulation including a second animated video depicting a second reconstruction of the accident;
providing the first animated video and the second animated video to an inconsistency detection model;
determining, via the inconsistency detection model, that one or more elements in the first animated video are inconsistent relative to one or more elements in the second animated video; and
automatically presenting, via an application, an alert indicating a high likelihood of unreliability associated with either or both of the first data and the second data.
2. The method of claim 1 , further comprising:
causing the first simulation to be presented via a user interface for the application;
receiving, via the user interface, a first user input for interacting with a first graphical element of the first simulation; and
presenting, in response to the first user input and via the user interface, additional details related to the first graphical element.
3. The method of claim 1 , wherein the second data includes one or more of vehicle telemetry for a vehicle involved in the accident, a police report of the accident, and image data showing aspects of the accident.
4. The method of claim 3 , wherein the determination that the first feature has a high likelihood of being inconsistent with information included in the second data is based on a comparison of synthetic data from the first simulation and synthetic data from the second simulation.
5. The method of claim 1 , further comprising:
generating a third simulation of the accident based on information extracted from both the first data and the second data, the third simulation including a greater number of graphical elements relative to either the first simulation and second simulation.
6. The method of claim 1 , wherein the first simulation includes an interactive 3D model of the accident.
7. The method of claim 1 , wherein the second simulation includes a different number of graphical elements relative to the first simulation.
8. The method of claim 1 , further comprising:
providing the first data to a predictive model;
determining, via the predictive model, predicted damage to a vehicle that should have resulted from the accident based on the information included in the first data; and
presenting, via the application, content describing or depicting the predicted damage.
9. The method of claim 8 , wherein the content includes a depiction of the predicted damage overlaid on a pictorial representation of a vehicle.
10. The method of claim 1 , further comprising:
providing the first data to a predictive model;
determining, via the predictive model, a predicted injury to a person that should have resulted from the accident based on the information included in the first data; and
presenting, via the application, content describing or depicting the predicted injury.
11. The method of claim 10 , wherein the content includes a depiction of the predicted injury overlaid on a pictorial representation of a person.
12. A system for generating a computer-implemented simulation of aspects of a vehicle accident, the system comprising a processor and machine-readable media including instructions which, when executed by the processor, cause the processor to:
receiving first data including a testimony of a witness to the accident;
receiving second data including information about the accident from a source other than the witness;
perform natural language processing (NLP) on the first data;
provide the processed first data to a keyword machine learning model to extract one or more terms detected in the first data that are associated with high degree of relevance for the accident;
feed the extracted terms to a generative engine to generate a first simulation of the accident based on elements represented by the extracted terms, wherein the first simulation includes a first animated video depicting a first reconstruction of the accident;
generate a second simulation of the accident based on information extracted from the second data, the second simulation including a second animated video depicting a second reconstruction of the accident;
provide the first animated video and the second animated video to an inconsistency detection model;
determine, via the inconsistency detection model, that one or more elements in the first animated video are inconsistent relative to one or more elements in the second animated video; and
automatically present, via an application, an alert indicating a high likelihood of unreliability associated with either or both of the first data and the second data.
13. The system of claim 12 , wherein the first simulation includes an interactive 3D model of the accident.
14. The system of claim 12 , wherein the instructions further cause the processor to:
provide the first data to a predictive model;
determine, via the predictive model, predicted damage to a vehicle that should have resulted from the accident based on the information included in the first data; and
present, via the application, content describing or depicting the predicted damage.
15. The system of claim 12 , wherein the instructions further cause the processor to:
provide the first data to a predictive model;
determine, via the predictive model, a predicted injury to a person that should have resulted from the accident based on the information included in the first data; and
present, via the application, content describing or depicting the predicted injury.
16. The system of claim 15 , wherein the content includes a depiction of the predicted injury overlaid on a pictorial representation of a person.
17. The system of claim 12 , wherein the second simulation includes a greater number of graphical elements relative to the first simulation.
18. The system of claim 12 , wherein the instructions further cause the processor to:
cause the first simulation to be presented via a user interface for the application;
receive, via the user interface, a first user input for interacting with a first graphical element of the first simulation; and
present, in response to the first user input and via the user interface, additional details related to the first graphical element.
19. The system of claim 12 , wherein the second data includes one or more of vehicle telemetry for a vehicle involved in the accident, a police report of the accident, and image data showing aspects of the accident.
20. The system of claim 14 , wherein the content includes a depiction of the predicted damage overlaid on a pictorial representation of a vehicle.