Entity learning recognition

1. A computer-implemented entity learning recognition training method, comprising:

in a training phase:

selecting a structural element from a plurality of structural elements, the plurality of structural elements being a shape that is associated with a feature of a face;

for one or more entities in a training database:

associating the structural element with one or more features in an aspect of the entities, the entities being the faces of the training database;

generating one or more augmented entities based on one or more combinations of the structural element and the features; and

adding the augmented entities to the training database; and

training a model based on a combination of original entities and the augmented entities in an augmented database,

wherein the entities comprise an image that is used for the training of the model, and

wherein the structural element comprises a visual element portion of the image with added noise.

2. The computer-implemented method of claim 1 , further comprising associating the visual element with an object is based on a geometric transform.

3. The computer-implemented method of claim 1 , wherein a generated augmented entity is based on n superimposition of the structural element to a sample entity.

4. The computer-implemented method of claim 1 , further comprising:

applying existing distortion resultants of the entity in the database to the generated augmented entity;

updating the database with the augmented entity or the combined augmented entity of the generated augmented entity and the existing distortion resultants; and

verifying if more generated augmented entities are required in the database to train the model based on a feedback from the model.

5. The computer-implemented method of claim 1 , wherein said generating, further comprises:

selecting a distortion model D(.), from a plurality of distortion models;

transforming an object image (F) as a function of a selected D(.);

to create an extended corpus of augmented object images {D(F)}; and

reconstructing the object image (F) from D(F), where reconstructed object image (F′)=R(D(F))+e, where R is a transform and e is a reconstruction error,

wherein the transforming the object image (F) further comprises selecting R based on the value of e, and

wherein said training the model is limited to one or more R that produces a reconstructed object image (F′) exceeding a threshold value.

6. The computer-implemented method of claim 1 , further comprising:

in an implementation phase:

applying a data point to the trained model to obtain a feedback of an efficiency of a correspondence between the data point and the trained model; and

wherein the feedback is used to update the model in the training phase by selecting additional entities and training the model with the additional generated augmented entities.

7. The computer-implement method of claim 1 , further comprising:

associating the visual element with an object is based on a geometric transformation;

applying existing distortion resultants of the entity in the database to the generated augmented entity;

updating the database with the augmented entity or the combined augmented entity of the generated augmented entity and the existing distortion resultants; and

verifying if more generated augmented entities are required in the database to train the model based on a feedback from the model,

wherein a generated augmented entity is based on n superimposition of the structural element to a sample entity.

8. A computer program product comprising a computer-readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause:

in a training phase:

selecting a structural element from a plurality of structural elements, the plurality of structural elements being a shape that is associated with a feature of a face;

for one or more entities in a training database:

associating the structural element with one or more features in an aspect of the entities, the entities being the faces of the training database;

generating one or more augmented entities based on one or more combinations of the structural element and the features; and

adding the augmented entities to the training database; and

training a model based on a combination of original entities and the augmented entities in an augmented database,

wherein the entities comprise an image that is used for the training of the model, and

wherein the structural element comprises a visual element portion of the image with added noise.

9. The computer program product of claim 8 , further comprising associating the visual element with object is based on a geometric transform.

10. The computer program product of claim 8 , wherein a generated augmented entity is based on n superimposition of the structural element to a sample entity.

11. The computer program product of claim 8 , further comprising:

applying existing distortion resultants of the entity in the database to the generated augmented entity;

updating the database with the augmented entity or the combined augmented entity of the generated augmented entity and the existing distortion resultants; and

verifying if more generated augmented entities are required in the database to train the model based on a feedback from the model.

12. The computer program product of claim 8 , wherein said generating, further comprises:

selecting a distortion model D(.), from a plurality of distortion models;

transforming an object image (F) as a function of a selected D(.);

to create an extended corpus of augmented object images {D(F)}; and

reconstructing the object image (F) from D(F), where reconstructed object image (F′)=R(D(F))+e, where R is a transform and e is a reconstruction error,

wherein the transforming the object image (F) further comprises selecting R based on the value of e, and

wherein said training the model is limited to one or more R that produces a reconstructed object image (F′) exceeding a threshold value.

13. The computer program product of claim 8 , further comprising:

in an implementation phase:

applying a data point to the trained model to obtain a feedback of an efficiency of a correspondence between the data point and the trained model; and

wherein the feedback is used to update the model in the training phase by selecting additional entities and training the model with the additional generated augmented entities.

14. An entity learning recognition system, said system comprising:

a processor; and

a memory, operably coupled to the processor, the memory storing instructions to cause the processor to perform:

in a training phase:

selecting a structural element from a plurality of structural elements, the plurality of structural elements being a shape that is associated with a feature of a face;

for one or more entities in a training database:

associating the structural element with one or more features in an aspect of the entities, the entities being the faces of the training database;

generating one or more augmented entities based on one or more combinations of the structural element and the features; and

adding the augmented entities to the training database; and

training a model based on a combination of original entities and the augmented entities in an augmented database,

wherein the entities comprise an image that is used for the training of the model, and

wherein the structural element comprises a visual element portion of the image with added noise.

15. The system of claim 14 , wherein said generating, further comprises:

selecting a distortion model D(.), from a plurality of distortion models;

transforming an object image (F) as a function of a selected D(.);

to create an extended corpus of augmented object images {D(F)}; and

reconstructing the object image (F) from D(F), where reconstructed object image (F′)=R(D(F))+e, where R is a transform and e is a reconstruction error,

wherein the transforming the object image (F) further comprises selecting R based on the value of e, and

wherein said training the model is limited to one or more R that produces a reconstructed object image (F′) exceeding a threshold value.

16. The system of claim 14 , embodied in a cloud-computing environment.