System and methods for determining a quality metric of three-dimensional image data

1 . A method comprising:

receiving, from user equipment, image data of an object;

generating, based at least in part on the image data, a three-dimensional mesh representing the object;

determining a percentage of the object represented by the three-dimensional mesh; and

determining, based at least in part on the percentage of the object, one or more quality metrics associated with the three-dimensional mesh.

2 . The method of claim 1 , further comprising:

determining an approximated ground plane associated with the three-dimensional mesh; and

aligning the three-dimensional mesh with the approximated ground plane.

3 . The method of any of claim 2 , wherein

determining, based at least in part on the alignment, a length, a width, and a girth of the mesh;

determining, based at least in part on the length, the width, the girth, and one or more threshold ranges a success of the alignment between the three-dimensional mesh and the approximated ground plane.

4 . The method of claim 1 , wherein the three-dimensional mesh is an initial three-dimensional mesh generated concurrently with generating the image data and the one or more quality scores are one or more initial quality score associated with the method further comprises:

generating, based at least in part on the image data and the initial three-dimensional mesh, a final three-dimensional mesh representing the object; and

determining, based at least in part on the final three-dimensional mesh, one or more final quality metrics associated with the final three-dimensional mesh.

5 . The method of claim 4 , wherein the initial three-dimensional mesh is generated using a truncated signed differences function (TSDF).

6 . The method of claim 5 , wherein determining the percentage of the object represented by the three-dimensional mesh is based at least in part on a representative object, the representative object generated by one or more machine learning models trained on image data and mesh data of a plurality of instances of the object in various positions, orientations, and under various conditions.

7 . The method of claim 1 , wherein determining the one or more quality metrics associated with the three-dimensional mesh is based at least in part on a geometric property of the three-dimensional mesh.

8 . The method of claim 7 , wherein the geometric property is one or more of the following:

skewness metric determined based on an equiangular skew of the three-dimensional mesh;

maximum angle metric determined based at least in part on a presence of first elements of the three-dimensional mesh with large angles within the three-dimensional mesh;

volume versus circumradius metric determined based at least in part on a quotient of an object volume and a radius of a circumscribed sphere of a representative model of the object;

volume versus length metric determined based at least in part on a quotient of an edge length of the three-dimensional mesh and a volume of the three-dimensional mesh compared to a representative model of the object;

conditions number metric determined based at least in part on one or more properties of a matrix transformation of the three-dimensional mesh to the representative model of the object;

growth rate metric determined based at least in part on a comparison of a size of a second element of the three-dimensional mesh to a size of neighboring elements of the three-dimensional mesh in one or more directions;

curved skewness metric determined based at least in part on a measure of a deformation of a third element of the three-dimensional mesh from a higher-order element associate with the third element of the three-dimensional mesh;

aspect ratio metric determined based at least in part on a similarity between a fourth element of the three-dimensional mesh and the fourth element within the representative model of the object; or

Jacobian metric determined based at least in part on a deviation of a fifth element of the three-dimensional mesh from the fifth element within the representative model of the object.

9 . The method of claim 1 , further comprising:

determining a number of holes within the three-dimensional mesh; and

wherein determining the one or more quality metrics associated with the three-dimensional mesh is based at least in part on the number of holes within the three-dimensional mesh.

10 . The method of claim 9 , further comprising:

detecting one or more holes within the three-dimensional mesh;

determining a size of individual ones of the one or more holes; and

classifying the individual ones of the one or more holes into a first category or a second category based at least in part on the corresponding size; and

wherein determining the one or more quality metrics associated with the three-dimensional mesh is based at least in part on the number of holes within the first category.

11 . The method of claim 9 , further comprising:

detecting one or more holes within the three-dimensional mesh;

determining two or more regions associated with the three-dimensional mesh; and

assigning individual ones of the one or more holes into one of the two or more regions; and

wherein determining the one or more quality metrics associated with the three-dimensional mesh is based at least in part on a number of holes in individual ones of the two or more regions.

12 . The method of claim 11 , wherein the two or more regions includes a first region and a second region and the method further comprises:

applying a first weight to individual holes assigned to the first region; and

applying a second weight to individual holes assigned to the second region; and

wherein determining the one or more quality metrics associated with the three-dimensional mesh is based at least in part on a metric associated with a weight associated with the one or more holes.

13 . One or more non-transitory computer-readable media storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

receiving, from user equipment, image data of an object;

generating, based at least in part on the image data, a three-dimensional mesh representing the object;

determining a percentage of the object represented by the three-dimensional mesh;

determining an approximated ground plane associated with the three-dimensional mesh;

aligning the three-dimensional mesh to the approximated ground plane;

detecting one or more holes associated with the three-dimensional mesh;

determining a size of individual ones of the one or more holes; and

determining, based at least in part on the percentage of the object, a number of the one or more holes, a size of the individual ones of the one or more holes, and a success of the alignment, one or more quality metrics associated with the three-dimensional mesh.

14 . The one or more non-transitory computer-readable media of claim 13 , further comprising:

determining a region of the three-dimensional mesh associated with the individual ones of the one or more holes; and

wherein determining the one or more quality metrics associated with the three-dimensional mesh is based at least in part on the region associated with the individual ones of the one or more holes.

15 . The one or more non-transitory computer-readable media of claim 13 , further comprising:

determining a category for individual ones of the one or more holes; and

wherein determining the one or more quality metrics associated with the three-dimensional mesh is based at least in part on the category associated with the individual ones of the one or more holes.

16 . The one or more non-transitory computer-readable media of claim 15 , wherein the category is determined, for the individual ones of the one or more holes, based at least in part on a size corresponding to each hole and one or more size thresholds.

17 . The one or more non-transitory computer-readable media of claim 13 , further comprising performing one or more operations based on the three-dimension mesh in response to the one or more quality metrics meeting or exceeding a threshold.

18 . The one or more non-transitory computer-readable media of claim 13 , further comprising preventing one or more operations based on the three-dimension mesh from initiating in response to the one or more quality metrics failing to meet or exceed a threshold.

19 . A system comprising:

one or more processors; and

one or more non-transitory computer-readable media storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:

receiving, from user equipment, image data of an object;

generating, based at least in part on the image data, a three-dimensional mesh representing the object;

determining a percentage of the object represented by the three-dimensional mesh;

detecting one or more holes associated with the three-dimensional mesh;

determining a size of individual ones of the one or more holes; and

determining, based at least in part on the percentage of the object and a size of the individual ones of the one or more holes, one or more quality metrics associated with the three-dimensional mesh.

20 . The system of claim 19 , wherein the operations further comprise:

determining a first region and a second region associated with the three-dimensional mesh; and

assigning the individual ones of the one or more holes to either the first region or the second region; and

wherein determining the one or more quality metrics associated with the three-dimensional mesh is based at least in part on a number of holes assigned to the first region.