Scene model enrichment using semantic labels

1. A method comprising:

at an electronic device including one or more processors and non-transitory memory:

obtaining a three-dimensional scene model of a physical environment including a plurality of points, wherein each of the plurality of points is associated with a set of coordinates in a three-dimensional space, wherein a subset of the plurality of points is associated with a particular cluster identifier and a particular semantic label;

retrieving a three-dimensional object model based on the particular semantic label, the three-dimensional object model including at least a plurality of points; and

updating the three-dimensional scene model by replacing the subset of the plurality of points with the three-dimensional object model.

2. The method of claim 1 , wherein obtaining the three-dimensional scene model includes:

obtaining an image of a physical environment;

obtaining a depth map of the image of the physical environment; and

generating the three-dimensional scene model based on the image of the physical environment and the depth map, wherein the plurality of points of the three-dimensional object model includes points corresponding to features of the physical environment not visible in the image of the physical environment.

3. The method of claim 1 , wherein obtaining the three-dimensional scene model includes:

obtaining a plurality of images of the physical environment from a plurality of different perspectives; and

generating the three-dimensional scene model based on the plurality of images of the physical environment, wherein the plurality of points of the three-dimensional object model includes points corresponding to features of the physical environment not visible in the plurality of images of the physical environment.

4. The method of claim 3 , further comprising displaying a representation of the physical environment from a virtual camera perspective in which representation of the features of the physical environment not visible in the plurality of images of the physical environment are displayed.

5. The method of claim 1 , further comprising displaying a representation of the physical environment including a representation of a first object of the physical environment and a representation of a second object of the physical environment corresponding to the three-dimensional object model, wherein the first object visually occludes at least a portion of the second object, wherein an occlusiveness of the representation of the first object is set such that the corresponding portion of the representation of the second object is displayed.

6. The method of claim 1 , wherein obtaining the three-dimensional scene model includes:

spatially disambiguating portions of the plurality of points of the three-dimensional scene model into a plurality of clusters including the subset of the plurality of points; and

determining the particular semantic label based on a volumetric arrangement of the points of the subset of the plurality of points.

7. The method of claim 6 , wherein spatially disambiguating portions of the plurality of points of the three-dimensional scene model into the plurality of clusters includes performing at least one of plane model segmentation or Euclidean cluster extraction.

8. The method of claim 6 , wherein determining the first semantic label includes providing the sets of three-dimensional coordinates of the points of the subset of the plurality of points to a machine-learning object classifier.

9. The method of claim 1 , further comprising altering the three-dimensional object model based on the subset of the plurality of points.

10. The method of claim 1 , wherein the three-dimensional object model includes the plurality of points of the three-dimensional object model as vertices of a mesh-based object model, wherein the mesh-based object model includes one or more edges between the vertices.

11. The method of claim 1 , wherein the three-dimensional object model further includes physics data associated with the plurality of points of the three-dimensional object model.

12. A device comprising:

a non-transitory memory; and

one or more processors to:

obtain a three-dimensional scene model of a physical environment including a plurality of points, wherein each of the plurality of points is associated with a set of coordinates in a three-dimensional space, wherein a subset of the plurality of points is associated with a particular cluster identifier and a particular semantic label;

retrieve a three-dimensional object model based on the particular semantic label, the three-dimensional object model including at least a plurality of points; and

update the three-dimensional scene model by replacing the subset of the plurality of points with the three-dimensional object model.

13. The device of claim 12 , wherein the one or more processors are to obtain the three-dimensional scene model by:

obtaining a plurality of images of the physical environment from a plurality of different perspectives; and

generating the three-dimensional scene model based on the plurality of images of the physical environment, wherein the plurality of points of the three-dimensional object model includes points corresponding to features of the physical environment not visible in the plurality of images of the physical environment.

14. The device of claim 13 , wherein the one or more processors are further to display a representation of the physical environment from a virtual camera perspective in which representation of the features of the physical environment not visible in the plurality of images of the physical environment are displayed.

15. The device of claim 12 , wherein the one or more processors are further to display a representation of the physical environment including a representation of a first object of the physical environment and a representation of a second object of the physical environment corresponding to the three-dimensional object model, wherein the first object visually occludes at least a portion of the second object, wherein an occlusiveness of the representation of the first object is set such that the corresponding portion of the representation of the second object is displayed.

16. The device of claim 12 , wherein the one or more processors are to obtain the three-dimensional scene model by:

spatially disambiguating portions of the plurality of points of the three-dimensional scene model into a plurality of clusters including the subset of the plurality of points; and

determining the particular semantic label based on a volumetric arrangement of the points of the subset of the plurality of points.

17. The device of claim 12 , wherein the one or more processors are further to alter the three-dimensional object model based on the subset of the plurality of points.

18. The device of claim 12 , wherein the three-dimensional object model includes the plurality of points of the three-dimensional object model as vertices of a mesh-based object model, wherein the mesh-based object model includes one or more edges between the vertices.

19. The device of claim 12 , wherein the three-dimensional object model further includes physics data associated with the plurality of points of the three-dimensional object model.

20. A non-transitory memory storing one or more programs, which, when executed by one or more processors of a device, cause the device to:

obtain a three-dimensional scene model of a physical environment including a plurality of points, wherein each of the plurality of points is associated with a set of coordinates in a three-dimensional space, wherein a subset of the plurality of points is associated with a particular cluster identifier and a particular semantic label;

retrieve a three-dimensional object model based on the particular semantic label, the three-dimensional object model including at least a plurality of points; and

update the three-dimensional scene model by replacing the subset of the plurality of points with the three-dimensional object model.