Generating and modifying representations of dynamic objects in an artificial reality environment

1. A method comprising, by a computing system:

receiving an image of a real environment using a camera worn by a user, the image of the real environment comprising an object of interest;

determining that a portion of the image of the real environment comprises the object of interest;

generating, based on the portion of the image of the real environment comprising the object of interest, a two-dimensional surface representing the object of interest;

receiving depth measurements of the real environment corresponding to the portion of the image of the real environment comprising the object of interest;

determining, based on the depth measurements, a depth of the two-dimensional surface representing the object of interest;

determining, based one or more previous images of the object of interest, a predicted pose of the object of interest at a time subsequent to when the image of the real environment was captured;

posing the two-dimensional surface in a coordinate system corresponding to the real environment based on the depth of the two-dimensional surface and the predicted pose of the object of interest;

determining a visibility of a virtual object relative to the object of interest by comparing a model of the virtual object with the two-dimensional surface; and

generating an output image based on the determined visibility of the virtual object.

2. The method of claim 1 , wherein determining the predicted pose of the object of interest is based on:

one or more movements of the camera worn by the user subsequent to a time the image of the real environment is captured; or

one or more movements of the object of interest prior to when the image of the real environment is captured.

3. The method of claim 2 , wherein the predicted pose of the object of interest is determined by one or more components of the computing system that are physically connected to a headset worn by the user.

4. The method of claim 1 , wherein determining the portion of the image of the real environment that comprises the object of interest comprises:

generating, using a neural network of a machine learning model, a bounding box encompassing the object of interest; and

outputting, using the bounding box, a segmentation mask associated with the object of interest.

5. The method of claim 4 , wherein the two-dimensional surface is based on the segmentation mask associated with the object of interest.

6. The method of claim 1 , wherein determining the depth of the two-dimensional surface representing the object of interest comprises:

removing one or more outliers from the depth measurements;

generating, after removing the one or more outliers, a histogram comprising the depth measurements; and

determining, using the histogram, a most-frequently observed value of the depth measurements.

7. The method of claim 1 , wherein the two-dimensional surface comprises a plurality of texels, wherein each texel is assigned a value that specifies a level of transparency that is to be accorded to that texel.

8. The method of claim 1 , wherein the depth measurements are calculated relative to a pose of the camera at a time the image of the real environment is captured.

9. The method of claim 1 , wherein the camera worn by the user is connected to a head-mounted display.

10. The method of claim 1 , wherein the two-dimensional surface is generated without accessing a three-dimensional model of the object of interest.

11. One or more computer-readable non-transitory storage media embodying software that is operable when executed by one or more processors to:

receive an image of a real environment using a camera worn by a user, the image of the real environment comprising an object of interest;

determine that a portion of the image of the real environment comprises the object of interest;

generate, based on the portion of the image of the real environment comprising the object of interest, a two-dimensional surface representing the object of interest;

receive depth measurements of the real environment corresponding to the portion of the image of the real environment comprising the object of interest;

determine, based on the depth measurements, a depth of the two-dimensional surface representing the object of interest;

determine, based one or more previous images of the object of interest, a predicted pose of the object of interest at a time subsequent to when the image of the real environment was captured;

pose the two-dimensional surface in a coordinate system corresponding to the real environment based on the depth of the two-dimensional surface and the predicted pose of the object of interest;

determine a visibility of a virtual object relative to the object of interest by comparing a model of the virtual object with the two-dimensional surface; and

generate an output image based on the determined visibility of the virtual object.

12. The one or more computer-readable non-transitory storage media of claim 11 , wherein determining the predicted pose of the object of interest is based on:

one or more movements of the camera worn by the user subsequent to a time the image of the real environment is captured; or

one or more movements of the object of interest prior to when the image of the real environment is captured.

13. The one or more computer-readable non-transitory storage media of claim 12 , wherein the predicted pose of the object of interest is determined by one or more components of a computing system that are physically connected to a headset worn by the user.

14. The one or more computer-readable non-transitory storage media of claim 11 , wherein determining the portion of the image of the real environment that comprises the object of interest comprises:

generating, using a neural network of a machine learning model, a bounding box encompassing the object of interest; and

outputting, using the bounding box, a segmentation mask associated with the object of interest.

15. The one or more computer-readable non-transitory storage media of claim 11 , wherein determining the depth of the two-dimensional surface representing the object of interest comprises:

removing one or more outliers from the depth measurements;

generating, after removing the one or more outliers, a histogram comprising the depth measurements; and

determining, using the histogram, a most-frequently observed value of the depth measurements.

16. A system comprising:

one or more processors; and

one or more computer-readable non-transitory storage media coupled to the one or more processors and comprising instructions operable when executed by the one or more processors to cause the system to:

receive an image of a real environment using a camera worn by a user, the image of the real environment comprising an object of interest;

determine that a portion of the image of the real environment comprises the object of interest;

generate, based on the portion of the image of the real environment comprising the object of interest, a two-dimensional surface representing the object of interest;

receive depth measurements of the real environment corresponding to the portion of the image of the real environment comprising the object of interest;

determine, based on the depth measurements, a depth of the two-dimensional surface representing the object of interest;

determine, based one or more previous images of the object of interest, a predicted pose of the object of interest at a time subsequent to when the image of the real environment was captured;

pose the two-dimensional surface in a coordinate system corresponding to the real environment based on the depth of the two-dimensional surface and the predicted pose of the object of interest;

determine a visibility of a virtual object relative to the object of interest by comparing a model of the virtual object with the two-dimensional surface; and

generate an output image based on the determined visibility of the virtual object.

17. The system of claim 16 , wherein determining the predicted pose of the object of interest is based on:

one or more movements of the camera worn by the user subsequent to a time the image of the real environment is captured; or

one or more movements of the object of interest prior to when the image of the real environment is captured.

18. The system of claim 17 , wherein the predicted pose of the object of interest is determined by one or more components of a computing system that are physically connected to a headset worn by the user.

19. The system of claim 16 , wherein determining the portion of the image of the real environment that comprises the object of interest comprises:

generating, using a first neural network of a machine learning model, a bounding box encompassing the object of interest; and

outputting, using the bounding box, a segmentation mask associated with the object of interest.

20. The system of claim 16 , wherein determining the depth of the two-dimensional surface representing the object of interest comprises:

removing one or more outliers from the depth measurements;

generating, after removing the one or more outliers, a histogram comprising the depth measurements; and

determining, using the histogram, a most-frequently observed value of the depth measurements.