Hierarchical context in risk assessment using machine learning

Patent metadata

Field	Value
Publication number	US-12488576-B2
Application number	US-202218075538-A
Country	US
Kind code	B2
Filing date	Dec 6, 2022
Priority date	Dec 7, 2021
Publication date	Dec 2, 2025
Grant date	Dec 2, 2025

Abstract

Methods, systems, and apparatus for receiving a request for a risk assessment for a parcel, receiving a set of images for the parcel, the set of images including two or more images, each image having an image scale and an image resolution that is different from other images in the set of images, providing a first-level feature embedding and a second-level feature embedding, the first-level feature embedding being provided by processing a first-level image through a first-level machine learning (ML) model, and the second-level feature embedding being provided by processing a second-level image through a second-level ML model, determining a risk assessment at least partially by processing each of the first-level feature embedding and a second-level feature embedding through a fusion network, and providing a representation of the risk assessment for display.

First claim

What is claimed is:

1 . A method performed by one or more processors, the method comprising:
receiving a request for a risk assessment for a parcel; receiving a set of images for the parcel, the set of images comprising two or more images, each image having (i) an image scale, (ii) an image resolution, and (iii) a set of features of the parcel captured within the image, being different from other images in the set of images; processing a first-level image of the set of images having a first image scale and a first image resolution through a first-level machine learning (ML) model to provide a first-level feature embedding, the first-level ML model being trained on a first set of images having the first image scale and the first image resolutions; processing a second-level image of the set of images having a second image scale and a second image resolution through a second-level ML model to provide a second-level feature embedding, the second-level ML model being trained on a second set of images having the second image scale and the second image resolution, the second image scale being different from the first image scale, the second image resolution being different from the first image resolution; determining a risk assessment at least partially by processing each of the first-level feature embedding and a second-level feature embedding through a fusion network; and providing a representation of the risk assessment for display.

2 . The method of claim 1 , wherein the first-level feature embedding comprises a parcel-level feature embedding that is generated at least partially based on a parcel-level image in the set of images, and the second-level feature embedding comprises one of a neighborhood-level feature embedding that is generated at least partially based on a neighborhood-level image in the set of images and a landscape-level feature embedding that is generated at least partially based on a landscape-level image in the set of images.

3 . The method of claim 2 , wherein the parcel-level image has the first image scale and the first image resolution, each being greater than the second image scale and the second image resolution of one of the neighborhood-level image and the landscape-level image.

4 . The method of claim 1 , wherein each image in the set of images represents the parcel within a threshold of time from a designated time associated with the request.

5 . The method of claim 1 , wherein at least one image in the set of images comprises an overhead view of the parcel and at least one other image in the set of images comprises one of a vegetation segmentation map and an elevation map.

6 . The method of claim 1 , further comprising providing a third-level feature embedding by processing a third-level image through a third-level ML model, wherein the third-level feature embedding is processed through the fusion network in determining the risk assessment.

7 . The method of claim 1 , wherein determining the risk assessment comprises converting a relative risk prediction to the risk assessment using a calibration curve.

8 . The method of claim 7 , wherein the relative risk prediction is output by a last layer of the fusion network.

9 . The method of claim 1 , wherein the first-level feature embedding is output from a non-final layer of the first-level ML model and the second-level feature embedding is output from a non-final layer of the second-level ML model.

10 . The method of claim 1 , wherein the first-level feature embedding is further provided by processing a set of first-level pixel data through the first-level ML model, and the second-level feature embedding is further provided by processing a set of second-level pixel data through the second-level ML model.

11 . A non-transitory computer storage medium encoded with a computer program, the computer program comprising instructions that when executed by a data processing apparatus cause the data processing apparatus to perform operations comprising:
receiving a request for a risk assessment for a parcel; receiving a set of images for the parcel, the set of images comprising two or more images, each image having (i) an image scale, (ii) an image resolution, and (iii) a set of features of the parcel captured within the image, being different from other images in the set of images; processing a first-level image of the set of images having a first image scale and a first image resolution through a first-level machine learning (ML) model to provide a first-level feature embedding, the first-level ML model being trained on a first set of images having the first image scale and the first image resolutions; processing a second-level image of the set of images having a second image scale and a second image resolution through a second-level ML model to provide a second-level feature embedding, the second-level ML model being trained on a second set of images having the second image scale and the second image resolution, the second image scale being different from the first image scale, the second image resolution being different from the first image resolution; determining a risk assessment at least partially by processing each of the first-level feature embedding and a second-level feature embedding through a fusion network; and providing a representation of the risk assessment for display.

12 . The non-transitory computer storage medium of claim 11 , wherein the first-level feature embedding comprises a parcel-level feature embedding that is generated at least partially based on a parcel-level image in the set of images, and the second-level feature embedding comprises one of a neighborhood-level feature embedding that is generated at least partially based on a neighborhood-level image in the set of images and a landscape-level feature embedding that is generated at least partially based on a landscape-level image in the set of images.

13 . The non-transitory computer storage medium of claim 12 , wherein the parcel-level image has the first image scale and the first image resolution, each being greater than the second image scale and the second image resolution of one of the neighborhood-level image and the landscape-level image.

14 . The non-transitory computer storage medium of claim 11 , wherein each image in the set of images represents the parcel within a threshold of time from a designated time associated with the request.

15 . The non-transitory computer storage medium of claim 11 , wherein at least one image in the set of images comprises an overhead view of the parcel and at least one other image in the set of images comprises one of a vegetation segmentation map and an elevation map.

16 . The non-transitory computer storage medium of claim 11 , wherein operations further comprise providing a third-level feature embedding by processing a third-level image through a third-level ML model, wherein the third-level feature embedding is processed through the fusion network in determining the risk assessment.

17 . The non-transitory computer storage medium of claim 11 , wherein determining the risk assessment comprises converting a relative risk prediction to the risk assessment using a calibration curve.

18 . The non-transitory computer storage medium of claim 17 , wherein the relative risk prediction is output by a last layer of the fusion network.

19 . The non-transitory computer storage medium of claim 11 , wherein the first-level feature embedding is output from a non-final layer of the first-level ML model and the second-level feature

Assignees

• X Dev Llc

Inventors

• Li Xin
• Gupta Akshina
• Singaraju Nishanth
• Cowan Eliot Julien
• Cowan Avery Noam

Classifications

• G06V20/38

  Outdoor scenes · CPC title

• G06V10/80

  Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level (multimodal speaker identification or verification G10L17/10) · CPC title

• G06V10/82

  using neural networks · CPC title

• G06V10/806 Primary

  of extracted features · CPC title

• G06V20/52 Primary

  Surveillance or monitoring of activities, e.g. for recognising suspicious objects (recognising microscopic objects G06V20/69) · CPC title