Detection of clarity markings in gemstones

1. A gemstone detection method comprising:

using a camera to generate a set of training pictures illustrating three-dimensional features of a gemstone cut with a pattern, each training picture in the set of training pictures comprising facets of and inclusions within the gemstone visible along a point-of-view (POV) through the gemstone;

generating a trained neural network by training an untrained neural network using the set of training pictures and a set of training sketches of edges of the facets of the gemstone generated from the set of training pictures such that, for each training picture, the trained neural network is capable of detecting facet edges to thereby produce machine-generated drawings approximating the facets of the gemstone in the training pictures and the training sketches the trained neural network was trained with;

using the trained neural network to iteratively the generate machine-generated drawings from the set of training pictures, wherein each of the machine-generated drawings identifies edges of the facets of the gemstone; and

combining the set of machine-generated drawings into a three-dimensional model.

2. The method according to claim 1 , wherein the gemstone comprises a diamond and the pattern is one of round, princess, oval, marquise, pear, cushion, emerald, asscher, radiant and heart.

3. The method according to claim 1 , wherein the POV is one or more of a top-down, sideways or bottom-up POV.

4. The method according to claim 1 , wherein the pictures of the set of training pictures comprise digital images.

5. The method according to claim 1 , wherein the generating of the trained neural network is invariant to carat weight and color of the gemstone.

6. The method according to claim 1 , further comprising fitting a template of a facet structure of a gemstone.

7. The method according to claim 1 , wherein the generating of the trained neural network comprises:

providing an untrained neural network with an input image of a gemstone;

registering that an output of the untrained neural network exists;

comparing, for each pixel in the input image, the output with a target output to generate an error signal for each pixel; and

iteratively repeating the providing, the registering and the comparing for multiple input images and multiple target outputs until the output of the untrained neural network approximates the target output.

8. The method according to claim 1 , wherein the generating of the trained neural network comprises training the untrained neural network to identify precise locations of inclusions within the gemstone and the method further comprises:

developing a plan to recut the gemstone around the precise locations to produce two or more secondary gemstones; and

re-cutting the gemstone into the secondary gemstones.

9. A computer program product stored on a non-transitory computer readable storage medium for gemstone analysis, comprising:

a processor; and

a memory unit having executable instructions stored thereon, which, when executed, cause the processor to execute a gemstone clarity marking detection method comprising:

using a camera to generate a set of training pictures illustrating three-dimensional features of a gemstone cut with a pattern, each training picture in the set of training pictures comprising facets of and inclusions within the gemstone visible along a point-of-view (POV) through the gemstone;

generating a trained neural network by training an untrained neural network using the set of training pictures and a set of training sketches of edges of the facets of the gemstone generated from the set of training pictures such that, for each training picture, the trained neural network is capable of detecting facet edges to thereby produce machine-generated drawings approximating the facets of the gemstone in the training pictures and the training sketches the trained neural network was trained with;

using the trained neural network to iteratively generate machine-generated drawings from the set of training pictures, wherein each of the machine-generated drawings identifies edges of the facets of the gemstone; and

combining the set of machine-generated drawings into a three-dimensional model.

10. The computer program product according to claim 9 , wherein the gemstone comprises a diamond and the pattern is one of round, princess, oval, marquise, pear, cushion, emerald, asscher, radiant and heart.

11. The computer program product according to claim 9 , wherein the POV is one or more of a top-down, sideways or bottom-up POV.

12. The computer program product according to claim 9 , wherein the pictures of the set of training pictures comprise digital images.

13. The computer program product according to claim 9 , wherein the generating of the trained neural network is invariant to carat weight and color of the gemstone.

14. The computer program product according to claim 9 , wherein the method further comprises fitting a template of a facet structure of a gemstone.

15. The computer program product according to claim 9 , wherein the generating of the trained neural network comprises:

providing an untrained neural network with an input image of a gemstone;

registering that an output of the untrained neural network exists;

comparing, for each pixel in the input image, the output with a target output to generate an error signal for each pixel; and

iteratively repeating the providing, the registering and the comparing for multiple input images and multiple target outputs until the output of the untrained neural network approximates the target output.

16. The computer program product according to claim 9 , wherein the generating of the trained neural network comprises training the untrained neural network to identify precise locations of inclusions within the gemstone and the method further comprises:

developing a plan to recut the gemstone around the precise locations to produce two or more secondary gemstones; and

re-cutting the gemstone into the secondary gemstones.

17. A gemstone analysis system comprising:

a camera;

a cutting tool;

a processor; and

a memory unit having executable instructions stored thereon, which, when executed, cause the processor to execute a gemstone clarity marking detection method comprising:

using a camera to generate a set of training pictures illustrating three-dimensional features of a gemstone cut with a pattern, each training picture in the set of training pictures comprising facets of and inclusions within the gemstone visible along a point-of-view (POV) through the gemstone;

generating a trained neural network by training an untrained neural network using the set of training pictures and a set of training sketches of edges of the facets of the gemstone generated from the set of training pictures such that, for each training picture, the trained neural network is capable of detecting facet edges to thereby produce machine-generated drawings approximating the facets of the gemstone in the training pictures and the training sketches the trained neural network was trained with; and

using the trained neural network to iteratively generate machine-generated drawings from the set of training pictures and to identify precise locations of inclusions within the gemstone,

wherein the generating of the trained neural network comprises providing an untrained neural network with an input image of a gemstone, registering that an output of the untrained neural network exists, comparing, for each pixel in the input image, the output with a target output to generate an error signal for each pixel, and iteratively repeating the providing, the registering and the comparing for multiple input images and multiple target outputs until the output of the untrained neural network approximates the target output, and

wherein the gemstone clarity marking detection method further comprises:

developing a plan to recut the gemstone around the precise locations to produce two or more secondary gemstones; and

controlling the cutting tool to re-cut the gemstone into the secondary gemstones.

18. The gemstone analysis system according to claim 17 , wherein the gemstone is a diamond and the cut pattern is one of round, princess, oval, marquise, pear, cushion, emerald, asscher, radiant and heart.

19. The gemstone analysis system according to claim 17 , wherein the POV is a top-down, sideways or bottom-up POV.

20. The gemstone analysis system according to claim 17 , wherein the pictures of the set of training pictures comprise digital images.