Face image quality recognition methods and apparatuses

1. A computer-implemented method for identity document face image quality recognition, comprising:

pairing, for each user of a plurality of users and to form a pair of face images, an identity document (ID) face image and a live face image;

generating, for each pair of face images and based on a face similarity between the ID face image and the live face image, a similarity score for the ID face image; and

training, based on ID face images and similarity scores corresponding to the ID face images, a model for ID face image quality recognition by:

ranking the similarity scores corresponding to the ID face images based on a quality of each of the ID face images;

dividing the ranked similarity scores into N percentiles, wherein N is greater than two; and

assigning N labels of ID face image quality corresponding to the N percentiles of the ranked similarity scores, wherein the N labels of ID face image quality represent a scale of a predefined range indicating qualities of the ID face images.

2. The computer-implemented method of claim 1 , wherein training the model for ID face image quality recognition comprises training a multiclass classifier.

3. The computer-implemented method of claim 2 , wherein training the multiclass classifier comprises:

dividing the ranked similarity scores into the N labels of ID face image quality;

assigning an ID face image with a corresponding label; and

training a multiclass classifier of N classes based on the ID face images and corresponding labels of the ID face images.

4. The computer-implemented method of claim 3 , wherein dividing the ranked similarity scores into the N labels of ID face image quality is percentile-based.

5. The computer-implemented method of claim 1 , wherein training a model for ID face image quality recognition comprises training a regression model.

6. The computer-implemented method of claim 1 , further comprising, prior to generating the similarity score for the ID face image:

determining, for each of the plurality of users, if the live face image meets one or more predetermined standards; and

excluding the live face image in response to determining that the live face image does not meet the one or more predetermined standards.

7. The computer-implemented method of claim 6 , wherein determining if the live face image meets the one or more predetermined standards comprises:

using a no-reference image quality model and/or a face pose estimation model to determine if the live face image meets the one or more predetermined standards; and

excluding the live face image in response to determining that the live face image has a low image quality or a non-frontal pose.

8. The computer-implemented method of claim 1 , wherein the similarity scores are normally distributed.

9. The computer-implemented method of claim 1 , wherein the model is a convolutional neural network (CNN)-based model.

10. The computer-implemented method of claim 1 , wherein the live face image comprises a face image collected during biometric authentication or collected during an electronic-Know Your Customer (eKYC) procedure.

11. A non-transitory computer-readable medium storing one or more instructions executable by a computer system to perform operations for identity document face image quality recognition, comprising:

pairing, for each user of a plurality of users and to form a pair of face images, an identity document (ID) face image and a live face image;

generating, for each pair of face images and based on a face similarity between the ID face image and the live face image, a similarity score for the ID face image; and

training, based on ID face images and similarity scores corresponding to the ID face images, a model for ID face image quality recognition by:

ranking the similarity scores corresponding to the ID face images based on a quality of each of the ID face images;

dividing the ranked similarity scores into N percentiles, wherein N is greater than two; and

assigning N labels of ID face image quality corresponding to the N percentiles of the ranked similarity scores, wherein the N labels of ID face image quality represent a scale of a predefined range indicating qualities of the ID face images.

12. The non-transitory computer-readable medium of claim 11 , wherein training the model for ID face image quality recognition comprises training a multiclass classifier.

13. The non-transitory computer-readable medium of claim 12 , wherein training the multiclass classifier comprises:

dividing the ranked similarity scores into the N labels of ID face image quality;

assigning an ID face image with a corresponding label; and

training a multiclass classifier of N classes based on the ID face images and corresponding labels of the ID face images.

14. The non-transitory computer-readable medium of claim 13 , wherein dividing the ranked similarity scores into the N labels of ID face image quality is percentile-based.

15. The non-transitory computer-readable medium of claim 11 , wherein training a model for ID face image quality recognition comprises training a regression model.

16. The non-transitory computer-readable medium of claim 11 , further comprising, prior to generating the similarity score for the ID face image:

determining, for each of the plurality of users, if the live face image meets one or more predetermined standards; and

excluding the live face image in response to determining that the live face image does not meet the one or more predetermined standards.

17. The non-transitory computer-readable medium of claim 16 , wherein determining if the live face image meets the one or more predetermined standards comprises:

using a no-reference image quality model and/or a face pose estimation model to determine if the live face image meets the one or more predetermined standards; and

excluding the live face image in response to determining that the live face image has a low image quality or a non-frontal pose.

18. The non-transitory computer-readable medium of claim 11 , wherein the similarity scores are normally distributed.

19. The non-transitory computer-readable medium of claim 11 , wherein the model is a convolutional neural network (CNN)-based model.

20. The non-transitory computer-readable medium of claim 11 , wherein the live face image comprises a face image collected during biometric authentication or collected during an electronic-Know Your Customer (eKYC) procedure.

21. A computer-implemented system for identity document face image quality recognition, comprising:

one or more computers; and

one or more computer memory devices interoperably coupled with the one or more computers and having tangible, non-transitory, machine-readable media storing one or more instructions that, when executed by the one or more computers, cause the one or more computers to perform one or more operations comprising:

pairing, for each user of a plurality of users and to form a pair of face images, an identity document (ID) face image and a live face image;

generating, for each pair of face images and based on a face similarity between the ID face image and the live face image, a similarity score for the ID face image; and

training, based on ID face images and similarity scores corresponding to the ID face images, a model for ID face image quality recognition by:

ranking the similarity scores corresponding to the ID face images based on a quality of each of the ID face images;

dividing the ranked similarity scores into N percentiles, wherein N is greater than two; and

assigning N labels of ID face image quality corresponding to the N percentiles of the ranked similarity scores, wherein the N labels of ID face image quality represent a scale of a predefined range indicating qualities of the ID face images.