Granted Patent B2

US 11,631,235 · App. 17/870,279 · Granted Apr 18, 2023

Inventors: Giacomo Vianello (Mountain View, CA); Peter Lorenzen (Mountain View, CA)

Assignee: Cape Analytics, Inc.

G06V10/273G06Q30/0278G06T7/11G06T7/60G06T2200/04

Abstract

Claims (25)

1. A method, comprising:

determining a measurement depicting a property component occluded by a set of occluding objects;

determining a parcel boundary associated with the property;

infilling an occluded portion of the property component within the measurement to generate an updated measurement depicting an unoccluded version of the property component, wherein the property component is infilled based on the parcel boundary; and

extracting a set of property attributes for the property from the updated measurement.

2. The method of claim 1 , wherein infilling the occluded portion of the property component comprises:

removing the set of occluding objects from the measurement; and

infilling removed portions of the measurement.

3. The method of claim 2 , wherein removing the set of occluding objects from the measurement comprises:

determining a segmentation mask for the set of occluding objects based on a color image of the property; and

masking the set of occluding objects from the measurement using the segmentation mask.

4. The method of claim 3 , wherein the measurement comprises a 3D measurement.

5. The method of claim 1 , wherein the measurement comprises a remote measurement.

6. The method of claim 1 , wherein the occluded portion of the property component is infilled using a generative model.

7. The method of claim 6 , wherein the generative model is trained by:

determining a set of ground-truth measurements depicting unoccluded property components;

simulating a set of training measurements by occluding portions of the property components within the set of ground-truth measurements; and

training the generative model to predict the ground-truth measurement from the respective training measurement by infilling the occluded portions.

8. The method of claim 1 , wherein the measurement is a 3D measurement, wherein infilling the occluded portion of the property component comprises infilling voxel values of the occluded portions.

9. The method of claim 1 , wherein the property component comprises a roof.

10. The method of claim 1 , wherein the set of property attributes is used as an input to an automated valuation model.

11. The method of claim 1 , wherein the set of property attributes is extracted from the updated measurement using a set of attribute models, wherein an attribute model of the set of attribute models further outputs an uncertainty value associated with a property attribute of the set of property attributes.

12. The method of claim 1 , wherein the property component is a roof and wherein the set of property attributes comprises a roof pitch, wherein extracting the roof pitch comprises:

determining a set of pitches from a roof segment within the updated measurement; and

determining the roof pitch based on the set of pitches.

Assignments (1)

Continuity (2)

Cited By (11)

US 12,229,845 US 12,243,301 US 12,293,579 US 12,333,788 US 12,518,395 US 12,561,936 US 12,567,242 US 12,579,786 US 12,700,042 US 12,718,488 US 12,743,844