Deep model matching methods for image transformation

1. A computerized method to automatically select a model among pre-training deep model library that matches pilot data for image transformation, the method comprising the steps of:

receiving pilot data and pre-trained deep model library including a set of pre-trained deep models into electronic storage means, wherein the pilot data is provided by a user;

performing a model matching scoring by computing means using the pilot data and the pre-trained deep model library to generate model matching score, wherein the model matching scoring is performed by evaluating a conformity of the pilot data and expected model input to generate an input conformity score, evaluating conformity of model output of the pilot data and expected model output to generate an output conformity score, and/or evaluating a model cycle consistency to the pilot data to generate a consistency score, and integrating the input conformity score, the output conformity score, and/or the consistency score to generate the model matching score; and

performing a model matching decision using the model matching score to generate a model matching decision output,

wherein the pilot data includes a representative image, its metadata, and application target specification.

2. The computerized method of claim 1 , wherein the model matching decision output is selected from a group consisting of (1) good model match, (2) no model match, and (3) undecided.

3. The computerized method of claim 1 , wherein the electronic storage means is in a cloud platform.

4. The computerized method of claim 1 , wherein the computing means is in a cloud platform.

5. The computerized method of claim 1 , wherein the image transformation is for microscopy image restoration.

6. The computerized method of claim 1 , wherein the image transformation is for microscopy image prediction.

7. A computerized method to progressively select a model among pre-training deep model library that matches pilot data for image transformation, the method comprising the steps of:

a) receiving pilot data and pre-trained deep model library including a set of pre-trained deep models into electronic storage means, wherein the pilot data is provided by a user;

b) performing a model matching scoring by computing means using the pilot data and the pre-trained deep model library to generate model matching score, wherein the model matching scoring is performed by evaluating a conformity of the pilot data and expected model input to generate an input conformity score, evaluating conformity of model output of the pilot data and expected model output to generate an output conformity score, and/or evaluating a model cycle consistency to the pilot data to generate a consistency score, and integrating the input conformity score, the output conformity score, and/or the consistency score to generate the model matching score;

c) performing a model matching decision using the model matching score to generate a model matching decision output selected from a group consisting of (1) good model match, (2) no model match, and (3) undecided;

d) if the model matching decision output from step c) is undecided, then if step c) has been repeated for a predetermined maximum number of times, setting the model matching decision output to no model match and terminating the method; or if step c) has not been repeated for the maximum number of times, inputting additional pilot data and adding the additional pilot data to the pilot data and repeating steps b) through c); and

e) if the model matching decision output from step c) is either good model match or no model match, terminating the method,

wherein the pilot data includes a representative image, its metadata and application target specification.

8. The computerized method of claim 7 , wherein the electronic storage means is in a cloud platform.

9. The computerized method of claim 7 , wherein the computing means is in a cloud platform.

10. The computerized method of claim 7 , wherein the image transformation is for microscopy image restoration.

11. The computerized method of claim 7 , wherein the image transformation is for microscopy image prediction.

12. A computerized method for deep adversarial model matching to automatically select a model among pre-training deep model library that matches pilot data for image transformation, the method comprising the steps of:

receiving pilot data and pre-trained deep adversarial model library into electronic storage means, wherein the pilot data is provided by a user;

performing a model matching scoring by computing means using the pilot data and the pre-trained deep adversarial model library to generate model matching score, wherein the model matching scoring is performed by evaluating a conformity of the pilot data and expected model input to generate an input conformity score, evaluating conformity of model output of the pilot data and expected model output to generate an output conformity score, and/or evaluating a model cycle consistency to the pilot data to generate a consistency score, and integrating the input conformity score, the output conformity score, and/or the consistency score to generate the model matching score; and

performing a model matching decision using the model matching score to generate a model matching decision output,

wherein the pilot data includes a representative image, its metadata and application target specification.

13. The computerized method of claim 12 , wherein the application target specification includes ground truth image, and the ground truth image is paired or un-paired.

14. The computerized method of claim 12 , wherein the pre-trained deep adversarial model library uses bi-directional generative adversarial network model.

15. The computerized method of claim 12 , wherein the model matching decision output is selected from a group consisting of (1) good model match, (2) no model match, and (3) undecided.

16. The computerized method of claim 12 , wherein the electronic storage means is in a cloud platform.

17. The computerized method of claim 12 , wherein the computing means is in a cloud platform.

18. The computerized method of claim 12 , wherein the image transformation is for microscopy image restoration.

19. The computerized method of claim 12 , wherein the image transformation is for microscopy image prediction.