Damage detection using machine learning

1. A method comprising:

receiving an image of at least a section of a vehicle;

processing the received image using damage detection neural network to detect a plurality of hail damage areas on the section of the vehicle and to classify each of the plurality of areas of damage according to the seriousness of the damage; wherein,

detecting a plurality of hail damage comprises:

detecting a plurality of damaged areas distributed over the entire section of the vehicle; and

differentiating the plurality of damaged areas from one or more areas of noise;

processing the received image using a further neural network to classify one or more sections of the vehicle as one or more panels of the vehicle bodywork; and

using the detected areas of damage, the classification of the seriousness of the damage and the classification of one or more panels to compute a panel damage density estimate.

2. The method according to claim 1 wherein the received image is selected from one of a monocular image or a stereo image.

3. The method according to claim 1 comprising, prior to processing the received image using the further neural network, generating an input to the further neural network from the received image by converting the received image to a binary image.

4. The method according to claim 3 comprising processing the converted image using a generator neural network to generate a modified image, wherein the generator neural network has been trained jointly with a discriminator neural network to generate modified images that have reduced image noise relative to input images to the generator neural network.

5. The method according to claim 1 wherein the plurality of areas of damage are one or more dents in the panels of the vehicle bodywork.

6. The method according to claim 5 wherein the seriousness of the damage is classified according to the depth and density of the dents.

7. The method according to claim 1 wherein the first neural network is trained on a data set comprising a mix of image formats, and wherein the further neural network is trained on a data set comprising a mix of image formats.

8. The method according to claim 7 wherein the mix of image formats include one or more 3D geometry files.

9. The method according to claim 8 wherein the one more 3D geometry files are augmented with simulated hail damage, the simulated hail damage simulated using hail impact analysis.

10. The method according to claim 1 wherein detecting a plurality of hail damage further comprises differentiating the plurality of damaged areas from one or more sources of noise.

11. The method according to claim 10 wherein the one or more sources of noise are selected from dust particles, dirt and specular reflection.

12. A system comprising:

one or more cameras;

one or more computing devices; and

one or more storage devices storing instructions that when executed by the one or more computers cause the one or more computers to perform operations comprising:

receiving an image of at least a section of a vehicle;

processing the received image using damage detection neural network to detect a plurality of hail damage areas on the section of the vehicle and to classify each of the plurality of areas of damage according to the seriousness of the damage; wherein,

detecting a plurality of hail damage comprises:

detecting a plurality of damaged areas distributed over the entire section of the vehicle; and

differentiating the plurality of damaged areas from one or more areas of noise;

processing the received image using a further neural network to classify one or more sections of the vehicle as one or more panels of the vehicle bodywork; and

using the detected areas of damage, the classification of the seriousness of the damage and the classification of one or more panels to compute a panel damage density estimate.

13. The system of claim 12 wherein the one or more cameras comprise at least two cameras and the image is a composite stereo 3D image generated from the at least two cameras.

14. The system of claim 12 wherein the one or more cameras comprise a cell-phone camera.

15. The system of claim 12 wherein the one or more cameras are located at a first location and the one or more computing and one or more storage devices are located at a second location.

16. One or more non-transitory computer-readable media storing instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:

receiving an image of at least a section of a vehicle;

processing the received image using damage detection neural network to detect a plurality of hail damage areas on the section of the vehicle and to classify each of the plurality of areas of damage according to the seriousness of the damage; wherein,

detecting a plurality of hail damage comprises:

detecting a plurality of damaged areas distributed over the entire section of the vehicle; and

differentiating the plurality of damaged areas from one or more areas of noise;

processing the received image using a further neural network to classify one or more sections of the vehicle as one or more panels of the vehicle bodywork; and

using the detected areas of damage, the classification of the seriousness of the damage and the classification of one or more panels to compute a panel damage density estimate.

17. A method comprising:

receiving an image of at least a section of a vehicle;

generating an input to a neural network from the received image by converting the received image to a binary image;

processing the converted image using a generator neural network to generate a modified image, wherein the generator neural network has been trained jointly with a discriminator neural network to generate modified images that have reduced image noise relative to input images to the generator neural network; and

processing the converted image using a further neural network to classify one or more sections of the vehicle as one or more panels of the vehicle bodywork.

18. The method of claim 17 wherein the received image is a CAD image.

19. The method of claim 18 wherein the one more CAD images are augmented with simulated hail damage, the simulated hail damage simulated using hail impact analysis.

20. The method of claim 17 further comprising using the processed image as input to a damage detection neural network.

21. The method of claim 17 wherein the wherein the modified image is used as a training input to one or more of a damaged detection neural network and a neural network to classify one or more sections of the vehicle.