IP Library Granted Patent US 11,508,051
Granted Patent B2
US 11,508,051 · App. 16/894,708 · Granted Nov 22, 2022

Image and data analystics model compatibility regulation methods

Inventors: Shih-Jong James Lee (Bellevue, WA); Hideki Sasaki (Bellevue, WA)
Assignee: LEICA MICROSYSTEMS CMS GMBH
G06T7/0002G06N20/00G06T2207/20081G06T2207/20084
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 11,508,051
App. No.
16/894,708
Granted
Nov 22, 2022
Kind
B2
Abstract

A computerized model compatibility regulation method for imaging applications first performs a target domain B application by computing means using at least one image X and target domain B image analytics to generate a target domain B application output for X. The method then applies a reference domain A application by computing means to generate reference domain A application output for X. The method further performs a compatibility assessment to generate at least one compatibility result for X. In addition, the method checks the compatibility result for X and if the check output is incompatible, the method performs online correction to generate a corrected application output for X.

Claims (52)

1. A computerized model compatibility regulation method for imaging applications, the method comprising the steps of:

a) inputting at least one image X and a target domain B image analytics into electronic storage means;

b) performing a target domain B application by computing means using the at least one image X and the target domain B image analytics to generate a target domain B application output for X;

c) performing a reference domain A application by computing means using the at least one image X to generate a reference domain A application output for X; and

d) performing a compatibility assessment by computing means using the target domain B application output for X and the reference domain A application output for X to generate at least one compatibility result for X.

2. The computerized model compatibility regulation method of claim 1 , wherein the reference domain A application is performed by a domain conversion referencing method.

3. The computerized model compatibility regulation method of claim 2 , wherein the domain conversion referencing method comprises a B to A domain conversion step and a domain A application step.

4. The computerized model compatibility regulation method of claim 3 , wherein the B to A domain conversion step uses a B to A domain converter to generate converted domain A image X′ and the domain A application step uses a converted domain A image analytics and the converted domain A image X′ to generate a domain A application output for X′ which is the reference domain A application output for X.

5. The computerized model compatibility regulation method of claim 4 , wherein the B to A domain converter is generated by performing a B to A domain converter training using the target domain B image analytics, a plurality of domain B training data and at least one domain A image.

6. The computerized model compatibility regulation method of claim 4 , wherein the converted domain A image analytics is generated by performing a B to A image analytics conversion training using a plurality of domain B training data and the B to A domain converter.

7. The computerized model compatibility regulation method of claim 1 , wherein the compatibility result for X comprises at least one difference metric.

8. The computerized model compatibility regulation method of claim 1 , further comprising the steps of:

e) performing a check using the at least one compatibility result for X to generate an output, that is compatible or incompatible; and

f) if the check output is incompatible, performing online correction to generate a corrected application output for X.

9. The computerized model compatibility regulation method of claim 8 , wherein the online correction step trains and applies a new domain converter to convert the at least one image X to match a target domain B.

10. A computerized multi-reference model compatibility regulation method for imaging applications, the method comprising the steps of:

a) inputting at least one image X and a target domain B analytics into electronic storage means;

b) performing a target domain B application by computing means using the at least one image X and the target domain B image analytics to generate a target domain B application output for X;

c) performing a plurality of reference domain Ai applications by computing means using the at least one image X to generate a plurality of reference domain Ai application outputs for X; and

d) performing a compatibility assessment by computing means using the target domain B application output for X and the plurality of reference domain Ai application outputs for X to generate at least one compatibility result for X.

11. The computerized multi-reference model compatibility regulation method of claim 10 , wherein the plurality of reference domain Ai applications are performed by a plurality of domain conversion referencing methods.

12. The computerized multi-reference model compatibility regulation method of claim 11 , wherein each of the domain conversion referencing methods comprises a B to Ai domain conversion step and a domain Ai application step.

13. The computerized multi-reference model compatibility regulation method of claim 12 , wherein the B to Ai domain conversion step uses a B to Ai domain converter to generate converted domain A image Xi and the domain Ai application step uses a converted domain Ai image analytics and the converted domain A image Xi to generate a domain Ai application output for Xi which is the reference domain Ai application output for X.

14. The computerized multi-reference model compatibility regulation method of claim 13 , wherein the B to Ai domain converter is generated by performing a B to Ai domain converter training using the target domain B image analytics, a plurality of domain B training data and at least one domain Ai image.

15. The computerized multi-reference model compatibility regulation method of claim 13 , wherein the converted domain Ai image analytics is generated by performing a B to Ai image analytics conversion training using a plurality of domain B training data and the B to Ai domain converter.

16. The computerized multi-reference model compatibility regulation method of claim 10 , wherein the compatibility result for X comprises at least one difference metric.

17. The computerized multi-reference model compatibility regulation method of claim 10 , further comprising the steps of:

e) performing a check using the at least one compatibility result for X to generate an output that is compatible or incompatible; and

f) if the check output is incompatible, performing online correction to generate a corrected application output for X.

18. The computerized multi-reference model compatibility regulation method of claim 17 , wherein the online correction step trains and applies a new domain converter to convert the at least one image X to match a target domain B.

19. A computerized model compatibility regulation method for data processing applications, the method comprising the steps of:

a) inputting at least one data X, a domain B analytics, a B to A data domain converter and a converted domain A analytics into electronic storage means;

b) performing a domain B application by computing means using the at least one data X and the domain B analytics to generate an application output for X;

c) performing B to A data domain conversion by computing means using the at least one data X and the B to A data domain converter to generate a converted domain A data X′;

d) performing a domain A application by computing means using the converted domain A data X′ and the converted domain A analytics to generate an application output for X′; and

e) performing a compatibility assessment by computing means using the application output for X and the application output for X′ to generate at least one compatibility result for X.

20. The computerized model compatibility regulation method for data processing applications of claim 19 , wherein the B to A data domain converter is generated by a B to A data domain matching converter training.

21. The computerized model compatibility regulation method for data processing applications of claim 20 , wherein the B to A data domain matching converter training comprises the steps of:

inputting a plurality of domain B training data and domain A data into electronic storage means,

performing a first phase domain matching converter training using the plurality of domain B training data and the domain B analytics to generate a phase 1 intermediate result composed of a phase 1 encoder E B and a phase 1 generator G B ; and

performing a second phase domain matching converter training using the plurality of domain B training data, the domain A data, the domain B analytics and the phase 1 intermediate result to simultaneously train and generate the B to A data domain converter containing encoders E A , E B and generators G A , G B .

22. The computerized model compatibility regulation method for data processing applications of claim 21 , wherein the step of performing B to A data domain conversion first applies E B to the at least one data X and then applies G A to generate the converted domain A data X′.

23. The computerized model compatibility monitoring method for data processing applications of claim 19 , wherein the converted domain A analytics is generated by a B to A data analytics conversion training.

24. The computerized model compatibility regulation method for data processing applications of claim 23 , wherein the B to A data analytics conversion training comprises the steps of:

inputting a plurality of domain B training data into electronic storage means;

performing a B to A data domain conversion by computing means using the plurality of domain B training data and the B to A data domain converter to generate converted domain A training data; and

performing an analytics training using the converted domain A training data to generate converted domain A analytics.

25. The computerized model compatibility regulation method for data processing applications of claim 19 , wherein the compatibility result for X comprises at least one difference metric.

26. The computerized model compatibility regulation method for data processing applications of claim 19 , further comprising the steps of:

f) performing a check using the compatibility result for X to generate an output that is compatible or incompatible; and

g) if the check output is incompatible, performing online correction to generate a corrected application output for X.

27. The computerized model compatibility regulation method for data processing applications of claim 26 , wherein the online correction step trains and applies a new domain converter to convert the at least one data X to match a target domain B.

Assignments (4)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 4, 2021
From: LEICA MICROSYSTEMS INC.
To: LEICA MICROSYSTEMS CMS GMBH
Reel/Frame 057697/0440 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Mar 16, 2021
From: SVISION LLC
To: LEICA MICROSYSTEMS INC.
Reel/Frame 055600/0752 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Dec 18, 2020
From: DRVISION TECHNOLOGIES LLC
To: SVISION LLC
Reel/Frame 054688/0279 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 8, 2020
From: LEE, SHIH-JONG JAMES; SASAKI, HIDEKI
To: DRVISION TECHNOLOGIES LLC
Reel/Frame 052870/0655 →