Damage detection from multi-view visual data

1. A method comprising:

determining an object model by applying a neural network to a plurality of images to identify a plurality of components of a designated object, each of the plurality of images being captured from a respective viewpoint, each of the plurality of images being selected from a video captured by a handheld recording device as it moves through space around the designated object, each of the images corresponding with one or more of a plurality of object model components, each of the plurality of object model components representing a respective region of the designated object;

determining respective component condition information for one or more of the plurality of object model components based on the plurality of images, the component condition information indicating a characteristic of damage incurred by the respective object portion corresponding with the object model component, wherein the characteristic of damage includes an estimated probability of damage to the respective object portion and an estimated severity of damage to the respective object portion, wherein the characteristic of damage is converted into a heatmap visual representation that shows the estimated probability of damage to the respective object portion and the estimated severity of damage to the respective object portion, wherein the heatmap is overlayed on a Multi-View Interactive Digital Media Representation (MVIDMR); and

storing the component condition information on a storage device.

2. The method recited in claim 1 , the method further comprising:

determining aggregated object condition information based on the component condition information, the aggregated object condition information indicating damage to the object as a whole.

3. The method recited in claim 2 , the method further comprising:

based on the aggregated object condition information, determining a standard view of the object that includes a visual representation of damage to the object.

4. The method recited in claim 3 , wherein the standard view of the object is selected from the group consisting of: a top-down view of the object, a multi-view representation of the object navigable in one or more directions, and a three-dimensional model of the object.

5. The method recited in claim 1 , wherein determining the object model includes estimating pose information for a designated one of the plurality of images, the pose information including a location and angle of a camera with respect to the designated object for the designated image.

6. The method recited in claim 5 , wherein determining the object model includes determining a three-dimensional skeleton of the designated object based on two-dimensional skeleton joints and the pose information.

7. The method recited in claim 6 , wherein the plurality of object model components are identified based at least in part on the three-dimensional skeleton of the designated object.

8. The method recited in claim 1 , wherein a designated one of the plurality of object model components corresponds with a designated subset of the images and a designated portion of the object, the method further comprising:

constructing a multi-view representation of the designated portion of the object at the handheld recording device based on the designated subset of the images, the multi-view representation being navigable in one or more directions.

9. The method recited in claim 1 , wherein the heatmap is overlayed on a multi-view representation stitched using images taken from different angles and locations.

10. The method recited in claim 1 , wherein determining the component condition information comprises:

applying a neural network to a subset of the images corresponding with the respective object model component.

11. The method recited in claim 10 , wherein the neural network receives as an input depth information captured from a depth sensor at the handheld recording device.

12. The method recited in claim 10 , wherein determining the component condition information further comprises:

aggregating neural network results computed for separate images corresponding with the respective object model component.

13. The method recited in claim 10 , the method further comprising:

constructing a multi-view representation of the designated object at the handheld recording device based on the plurality of images, the multi-view representation being navigable in one or more directions.

14. The method recited in claim 10 , wherein the object is a vehicle, and wherein the object model includes a three-dimensional skeleton of the vehicle, and wherein the plurality of object model components include each of a left vehicle door, a right vehicle door, and a windshield.

15. The method recited in claim 10 ,

determining that the plurality of images exhibit a coverage level that is beneath a designated threshold; and

providing a recording guidance instruction to capture additional image data.

16. The method recited in claim 15 , wherein the plurality of images are captured at a first level of clarity or detail, and wherein the recording guidance instruction indicates a second level of clarity or detail.

17. One or more non-transitory computer readable media having instructions stored thereon for performing a method, the method comprising:

determining an object model by applying a neural network to a plurality of images to identify a plurality of components of a designated object, each of the plurality of images being captured from a respective viewpoint, each of the plurality of images being selected from a video captured by a handheld recording device as it moves through space around the designated object, each of the images corresponding with one or more of a plurality of object model components, each of the plurality of object model components representing a respective region of the designated object;

determining respective component condition information for one or more of the plurality of object model components based on the plurality of images, the component condition information indicating a characteristic of damage incurred by the respective object portion corresponding with the object model component, wherein the characteristic of damage includes an estimated probability of damage to the respective object portion and an estimated severity of damage to the respective object portion, wherein the characteristic of damage is converted into a heatmap visual representation that shows the estimated probability of damage to the respective object portion and the estimated severity of damage to the respective object portion, wherein the heatmap is overlayed on a Multi-View Interactive Digital Media Representation (MVIDMR); and

storing the component condition information on a storage device.

18. A mobile computing device comprising:

a camera configured to capture a video of a designated object as it moves through space around the designated object;

a processor configured to:

select a plurality of images of the designated object from the video, each of the plurality of images being captured from a respective viewpoint,

determine an object model by applying a neural network to the plurality of images to identify a plurality of components of the designated object, each of the images corresponding with one or more of a plurality of object model components, each of the plurality of object model components representing a respective region of the designated object, and

determine respective component condition information for one or more of the plurality of object model components based on the plurality of images, the component condition information indicating a characteristic of damage incurred by the respective object portion corresponding with the object model component, wherein the characteristic of damage includes an estimated probability of damage to the respective object portion and an estimated severity of damage to the respective object portion, wherein the characteristic of damage is converted into a heatmap visual representation that shows the estimated probability of damage to the respective object portion and the estimated severity of damage to the respective object portion, wherein the heatmap is overlayed on a Multi-View Interactive Digital Media Representation (MVIDMR); and

a storage device configured to store the component condition information.

19. The mobile computing device recited in claim 18 , wherein the processor is further configured to:

determine aggregated object condition information based on the component condition information, the aggregated object condition information indicating damage to the object as a whole; and

based on the aggregated object condition information, determine a standard view of the object that includes a visual representation of damage to the object.