Digital garment generation

1. A method comprising, by a computing device:

generating a front panel of a garment based on one or more images comprising the garment;

generating a back panel of the garment;

aligning the front panel and the back panel in a three-dimensional space so that the front panel is in front of a three-dimensional body and the back panel is behind the three-dimensional body;

identifying one or more pairs of boundary segments of the front panel and the back panel, wherein each pair of boundary segments of the front panel and the back panel are to be attached together;

generating a digital garment by attaching each of the identified one or more pairs of boundary segments of the front panel and the back panel through a plurality of iterative simulations using a physics simulation model;

detecting a user's action to a piece of the digital garment using real-time hand tracking techniques; and

deforming the digital garment to meet additional constraints caused by the user's action by providing information associated with the detection to a physics simulation model.

2. The method of claim 1 , wherein generating the front panel comprising:

accessing the one or more images comprising a front of the garment;

generating a front-segmentation mask identifying the front of the garment in the one or more accessed images; and

generating the front panel corresponding to the generated front-segmentation mask.

3. The method of claim 2 , wherein the front-segmentation mask is two-dimensional.

4. The method of claim 1 , wherein the back panel is a duplicate of the front panel.

5. The method of claim 1 , wherein the back panel is generated by:

accessing one or more second images comprising a back of the garment;

generating a back-segmentation mask identifying the back of the garment in the one or more second images; and

generating the back panel corresponding to the generated back-segmentation mask.

6. The method of claim 1 , wherein the front panel and the back panel comprise three-dimensional flat meshes.

7. The method of claim 1 , wherein the three-dimensional body is a three-dimensional template of a human body.

8. The method of claim 1 , wherein the three-dimensional body is generated based on a three-dimensional scan or a depth map of a user wearing the garment.

9. The method of claim 1 , wherein identifying the one or more pairs of boundary segments of the front panel and the back panel comprises:

drawing virtual lines that perpendicularly intersect with boundaries of the front panel and the back panel; and

identifying each continuous portion of the boundaries whose corresponding virtual lines do not intersect with the three-dimensional body.

10. The method of claim 1 , wherein each iteration of the plurality of iterative simulations comprises:

decreasing a distance between each pair of boundary segments of the front panel and the back panel; and

deforming the garment based on the distance between each pair of boundary segments of the front panel and the back panel and a shape of the three-dimensional body.

11. The method of claim 1 , further comprising:

determining that the three-dimensional body moves from the first pose to a second pose; and

deforming the digital garment according to the second pose of the three-dimensional body.

12. The method of claim 1 , wherein the generated digital garment is used for a virtual try-on, in which a user drapes the digital garment to an avatar representing the user.

13. The method of claim 12 , wherein detecting the user's action to the piece of the digital garment comprises observing collisions between meshes corresponding to the avatar and meshes corresponding to the digital garment.

14. The method of claim 12 , wherein detecting the user's action to the piece of the digital garment comprises:

detecting that a distance between fingers of the avatar and the digital garment is less than a threshold; and

generating a three-dimensional volume force field around the fingers, in which an amount of force at a point within the force field depends on a distance from the fingers, wherein any portion of the digital garment within the force field is influenced by the force.

15. The method of claim 12 , wherein detecting the user's action to the piece of the digital garment comprises:

generating a point cloud of a garment the user is wearing using depth sensing techniques; and

providing the point cloud to a physics simulation model that keeps the digital garment to be consistent with the point cloud.

16. The method of claim 12 , wherein the user's action comprises touching, pinching, or holding.

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

generate a front panel of a garment based on one or more images comprising the garment;

generate a back panel of the garment;

align the front panel and the back panel in a three-dimensional space so that the front panel is in front of a three-dimensional body and the back panel is behind the three-dimensional body;

identify one or more pairs of boundary segments of the front panel and the back panel, wherein each pair of boundary segments of the front panel and the back panel are to be attached together;

generate a digital garment by attaching each of the identified one or more pairs of boundary segments of the front panel and the back panel through a plurality of iterative simulations using a physics simulation model;

detect a user's action to a piece of the digital garment using real-time hand tracking techniques; and

deform the digital garment to meet additional constraints caused by the user's action by providing information associated with the detection to a physics simulation model.

18. The media of claim 17 , wherein generating the front panel comprising:

accessing the one or more images comprising a front of the garment;

generating a front-segmentation mask identifying the front of the garment in the one or more accessed images; and

generating the front panel corresponding to the generated front-segmentation mask.

19. A system comprising:

one or more processors; and

a non-transitory memory coupled to the processors comprising instructions executable by the processors, the processors operable when executing the instructions to:

generate a front panel of a garment based on one or more images comprising the garment;

generate a back panel of the garment;

align the front panel and the back panel in a three-dimensional space so that the front panel is in front of a three-dimensional body and the back panel is behind the three-dimensional body;

identify one or more pairs of boundary segments of the front panel and the back panel, wherein each pair of boundary segments of the front panel and the back panel are to be attached together;

generate a digital garment by attaching each of the identified one or more pairs of boundary segments of the front panel and the back panel through a plurality of iterative simulations using a physics simulation model;

detect a user's action to a piece of the digital garment using real-time hand tracking techniques; and

deform the digital garment to meet additional constraints caused by the user's action by providing information associated with the detection to a physics simulation model.